ckermit70.txt

   The Kermit Project
   Columbia University
   612 West 115th Street
   New York NY 10025 USA
   kermit@columbia.edu
   ...since 1981

Supplement to Using C-Kermit, 2nd Edition

For C-Kermit 7.0

As of C-Kermit version:  7.0.196
This file created:  8 February 2000
This file last updated:
Mon Aug  8 10:39:18 2011


Authors: Frank da Cruz and Christine M. Gianone
Address: The Kermit Project
         Columbia University
         612 West 115th Street
         New York NY 10025-7799
         USA
Fax:     +1 (212) 662-6442
E-Mail: kermit-support@columbia.edu
Web:    http://www.columbia.edu/kermit/
Or:     http://www.kermit-project.org/
Or:     http://www.columbia.nyc.ny.us/kermit/

NOTICES

   This document:
          Copyright © 1997, 2000, Frank da Cruz and Christine M. Gianone.
          All rights reserved.

   Kermit 95:
          Copyright © 1995, 2000, Trustees of Columbia University in the
          City of New York. All rights reserved.

   C-Kermit:
          Copyright © 1985, 2000,
          Trustees of Columbia University in the City of New York. All
          rights reserved. See the C-KermitCOPYING.TXT file or the
          copyright text in theckcmai.c module for disclaimer and
          permissions.

   When Kerberos(TM) and/or SRP(TM) (Secure Remote Password) and/or SSL
          protocol are included:
          Portions Copyright © 1990, Massachusetts Institute of
          Technology.
          Portions Copyright © 1991, 1993 Regents of the University of
          California.
          Portions Copyright © 1991, 1992, 1993, 1994, 1995 by AT&T.
          Portions Copyright © 1997, Stanford University.
          Portions Copyright © 1995-1997, Eric Young <eay@cryptosoft.com>.

   For the full text of the third-party copyright notices, see
  Appendix V.

WHAT IS IN THIS FILE

   This file lists changes made to C-Kermit since the second edition of
   the bookUsing C-Kermit was published and C-Kermit 6.0 was released
   in November 1996. Use this file as a supplement to the second edition
   of Using C-Kermit until the third edition is published some time in
   2000. If the "most recent update" shown above is long ago, contact
   Columbia University to see if there is a newer release.

   For further information, also see theCKCBWR.TXT ("C-Kermit
   beware") file for hints, tips, tricks, restrictions, frequently asked
   questions, etc, plus the system-specific "beware file", e.g.
  CKUBWR.TXT for UNIX,CKVBWR.TXT for VMS, etc, and also any
   system-specific update files such as KERMIT95.HTM for Kermit 95 (in the
   DOCS\MANUAL\ subdirectory of your K95 directory).

     This Web-based copy of the C-Kermit 7.0 update notes supersedes the
     plain-text CKERMIT2.TXT file. All changes after 19 January 2000
     appear only here in the Web version. If you need an up-to-date
     plain-text copy, use your Web browser to save this page as plain
     text.

ABOUT FILENAMES

   In this document, filenames are generally shown in uppercase, but on
   file systems with case-sensitive names such as UNIX, OS-9, and AOS/VS,
   lowercase names are used:ckubwr.txt,ckermit70.txt, etc.

ADDITIONAL FILES

   Several other files accompany this new Kermit release:

   SECURITY.TXT
          Discussion of Kermit's new authentication and encryption
          features:

          +Plain-text version
          +HTML (hypertext) version

   IKSD.TXT
          How to install and manage an Internet Kermit Service Daemon.

          +Plain-text version
          +HTML (hypertext) version

          Also seecuiksd.htm for instructions for use.

   TELNET.TXT
          A thorough presentation of Kermit's new advanced Telnet features
          and controls.

          +Plain-text version
          +HTML (hypertext) version

THE NEW C-KERMIT LICENSE

   The C-Kermit license was rewritten for version 7.0 to grant automatic
   permission to packagers of free operating-system distributions to
   include C-Kermit 7.0. Examples include Linux (GNU/Linux), FreeBSD,
   NetBSD, etc. The new license is in theCOPYING.TXT file, and is
   also displayed by C-Kermit itself when you give the VERSION or
   COPYRIGHT command. The new C-Kermit license does not apply to
  Kermit 95.

ACKNOWLEDGMENTS

   Thanks to Jeff Altman, who joined the Kermit Project in 1995, for much
   of what you see in C-Kermit 7.0, especially in the networking and
   security areas, and his key role in designing and implementing the
   Internet Kermit Service Daemon. And special thanks to Lucas Hart for
   lots of help with the VMS version; to Peter Eichhorn for continuous
   testing on the full range of HP-UX versions and for a consolidated set
   of HP-UX makefile targets; and to Colin Allen, Mark Allen, Roger Allen,
   Ric Anderson, William Bader, Mitch Baker, Mitchell Bass, Nelson Beebe,
   Gerry Belanger, Jeff Bernsten, Mark Berryman, John Bigg, Volker
   Borchert, Jonathan Boswell, Tim Boyer, Frederick Bruckman, Kenneth
   Cochran, Jared Crapo, Bill Delaney, Igor Sobrado Delgado, Clarence
   Dold, Joe Doupnik, John Dunlap, Max Evarts, Patrick French, Carl
   Friedberg, Carl Friend, Hirofumi Fujii, Andrew Gabriel, Gabe Garza,
   Boyd Gerber, David Gerber, George Gilmer, Hunter Goatley, DJ Hagberg,
   Kevin Handy, Andy Harper, Randolph Herber, Sven Holström, Michal
   Jaegermann, Graham Jenkins, Dick Jones, Terry Kennedy, Robert D Keys,
   Nick Kisseberth, Igor Kovalenko, David Lane, Adam Laurie, Jeff
   Liebermann, Eric Lonvick, Hoi Wan Louis, Arthur Marsh, Gregorie Martin,
   Peter Mauzey, Dragan Milicic, Todd Miller, Christian Mondrup, Daniel
   Morato, Dat Nguyen, Herb Peyerl, Jean-Pierre Radley, Steve Rance,
   Stephen Riehm, Nigel Roles, Larry Rosenman, Jay S Rouman, David
   Sanderson, John Santos, Michael Schmitz, Steven Schultz, Bob Shair,
   Richard Shuford, Fred Smith, Michael Sokolov, Jim Spath, Peter Szell,
   Ted T'so, Brian Tillman, Linus Torvalds, Patrick Volkerding, Martin
   Vorländer, Steve Walton, Ken Weaverling, John Weekley, Martin Whitaker,
   Jim Whitby, Matt Willman, Joellen Windsor, Farrell Woods, and many
   others for binaries, hosting, reviews, suggestions, advice, bug
   reports, and all the rest over the 3+ year C-Kermit 7.0 development
   cycle. Thanks to Russ Nelson and the board of the Open Software
   Initiative (http://www.opensource.org) for their cooperation in
   developing the new C-Kermit license and to the proprietors of those
   free UNIX distributions that have incorporated C-Kermit 7.0 for their
   cooperation and support, especially FreeBSD's Jörg Wunsch.

NOTE TO KERMIT 95 USERS

   Kermit 95 and C-Kermit share the same command and scripting language,
   the same Kermit file-transfer protocol implementation, and much else
   besides.

   Like the bookUsing C-Kermit, this file concentrates on the aspects
   of C-Kermit that are common to all versions: UNIX, VMS, Windows, OS/2,
   VOS, AOS/VS, etc. Please refer to your Kermit 95 documentation for
   information that is specific to Kermit 95.

   C-Kermit 7.0 corresponds to Kermit 95 1.1.19.

C-KERMIT VERSIONS AND VERSION NUMBERS

   "C-Kermit" refers to all the many programs that are compiled in whole
   or in part from common C-language source code, comprising:

     * A Kermit file transfer protocol module
     * A command parser and script execution module
     * A modem-dialing module
     * A network support module
     * A character-set translation module.

   and several others. These "system-independent" modules are combined
   with system-dependent modules for each platform to provide the required
   input/output functions, and also in some cases overlaid with an
   alternative user interface, such as Macintosh Kermit's point-and-click
   interface, and in some cases also a terminal emulator, as Kermit 95.

   The C-Kermit version number started as 1.0, ... 3.0, 4.0, 4.1 and then
   (because of confusion at the time with Berkeley UNIX 4.2), 4B, 4C, and
   so on, with the specific edit number in parentheses, for example
   4E(072) or 5A(188). This scheme was used through 5A(191), but now we
   have gone back to the traditional numbering scheme with decimal points:
   major.minor.edit; for example 7.0.196. Internal version numbers (the
   \v(version) variable), however, are compatible in C-Kermit 5A upwards.

   Meanwhile, C-Kermit derivatives for some platforms (Windows, Macintosh)
   might go through several releases while C-Kermit itself remains the
   same. These versions have their own platform-specific version numbers,
   such as Kermit 95 1.1.1, 1.1.2, and so on.

   C-Kermit Version History:

  1.0       1981-1982         Command-line only, 4.2 BSD UNIX only
  2.0       (*)               (who remembers...)
  3.0       May 1984          Command-line only, supports several platforms
  4.0-4.1   Feb-Apr 1985 (*)  First interactive and modular version
  4C(050)   May 1985
  4D(060)   April 1986
  4E(066)   August 1987       Long packets
  4E(068)   January 1988
  4E(072)   January 1989
  4F(095)   August 1989 (*)   Attribute packets
  5A(188)   November 1992     Scripting, TCP/IP, sliding windows (1)
  5A(189)   September 1993    Control-char unprefixing
  5A(190)   October 1994      Recovery
  5A(191)   April 1995        OS/2 only
  6.0.192   September 1996    Intelligent dialing, autodownload, lots more (2)
  6.1.193   1997-98 (*)       Development only
  6.1.194   June 1998         K95 only - switches, directory recursion, more
  7.0.195   August 1999       IKSD + more (CU only as K95 1.1.18-CU)
  7.0.196   1 January 2000    Unicode, lots more

   (*) Never formally released (4.0 was a total rewrite)
   (1) Using C-Kermit, 1st Edition
   (2) Using C-Kermit, 2nd Edition

CONTENTS

 I. C-KERMIT DOCUMENTATION

 II. NEW FEATURES

     (0) INCOMPATIBILITIES WITH PREVIOUS RELEASES
     (1) PROGRAM AND FILE MANAGEMENT AND COMMANDS
         1.0. Bug fixes
         1.1. Command Continuation
         1.2. Editor Interface
         1.3. Web Browser and FTP Interface
         1.4. Command Editing
         1.5. Command Switches
               1.5.1. General Switch Syntax
               1.5.2. Order and Effect of Switches
               1.5.3. Distinguishing Switches from Other Fields
               1.5.4. Standard File Selection Switches
               1.5.5. Setting Preferences for Different Commands
         1.6. Dates and Times
         1.7. Partial Completion of Keywords
         1.8. Command Recall
         1.9. EXIT Messages
         1.10. Managing Keyboard Interruptions
         1.11. Taming the Wild Backslash -- Part Deux
               1.11.1. Background
               1.11.2. Kermit's Quoting Rules
               1.11.3. Passing DOS Filenames from Kermit to Shell Commands
               1.11.4. Using Variables to Hold DOS Filenames
               1.11.5. Passing DOS Filenames as Parameters to Macros
               1.11.6. Passing DOS File Names from Macro Parameters to
			           the DOS Shell
               1.11.7. Passing DOS Filenames to Kermit from the Shell
         1.12. Debugging
         1.13. Logs
         1.14. Automatic File-Transfer Packet Recognition at
		       the Command Prompt
         1.15. The TYPE Command
         1.16. The RESET Command
         1.17. The COPY and RENAME Commands
         1.18. The MANUAL Command
         1.19. String and Filename Matching Patterns
         1.20. Multiple Commands on One Line
         1.21. What Do I Have?
         1.22. Generalized File Input and Output
               1.22.1. Why Another I/O System?
               1.22.2. The FILE Command
               1.22.3. FILE Command Examples
               1.22.4. Channel Numbers
               1.22.5. FILE Command Error Codes
               1.22.6. File I/O Variables
               1.22.7. File I/O Functions
               1.22.8. File I/O Function Examples
         1.23. The EXEC Command
         1.24. Getting Keyword Lists with '?'
     (2) MAKING AND USING CONNECTIONS
         2.0. SET LINE and SET HOST Command Switches
         2.1. Dialing
              2.1.1. The Dial Result Message
              2.1.2. Long-Distance Dialing Changes
              2.1.3. Forcing Long-Distance Dialing
              2.1.4. Exchange-Specific Dialing Decisions
              2.1.5. Cautions about Cheapest-First Dialing
              2.1.6. Blind Dialing (Dialing with No Dialtone)
              2.1.7. Trimming the Dialing Dialog
              2.1.8. Controlling the Dialing Speed
              2.1.9. Pretesting Phone Number Conversions
              2.1.10. Greater Control over Partial Dialing
              2.1.11. New DIAL-related Variables and Functions
              2.1.12. Increased Flexibility of PBX Dialing
              2.1.13. The DIAL macro - Last-Minute Phone Number Conversions
              2.1.14. Automatic Tone/Pulse Dialing Selection
              2.1.15. Dial-Modifier Variables
              2.1.16. Giving Multiple Numbers to the DIAL Command
         2.2. Modems
              2.2.1. New Modem Types
              2.2.2. New Modem Controls
         2.3.TELNET and RLOGIN
              2.3.0. Bug Fixes
              2.3.1. Telnet Binary Mode Bug Adjustments
              2.3.2. VMS UCX Telnet Port Bug Adjustment
              2.3.3. Telnet New Environment Option
              2.3.4. Telnet Location Option
              2.3.5. Connecting to Raw TCP Sockets
              2.3.6. Incoming TCP Connections
         2.4. The EIGHTBIT Command
         2.5. The Services Directory
         2.6. Closing Connections
         2.7. Using C-Kermit with External Communication Programs
              2.7.0. C-Kermit over tn3270 and tn5250
              2.7.1. C-Kermit over Telnet
              2.7.2. C-Kermit over Rlogin
              2.7.3. C-Kermit over Serial Communication Programs
              2.7.4. C-Kermit over Secure Network Clients
              2.7.4.1. SSH
              2.7.4.2. SSL
              2.7.4.3. SRP
              2.7.4.4. SOCKS
              2.7.4.5. Kerberos and SRP
         2.8. Scripting Local Programs
         2.9. X.25 Networking
              2.9.1. IBM AIXLink/X.25 Network Provider Interface for AIX
              2.9.2. HP-UX X.25
         2.10. Additional Serial Port Controls
         2.11. Getting Access to the Dialout Device
         2.12. The Connection Log
         2.13. Automatic Connection-Specific Flow Control Selection
         2.14. Trapping Connection Establishment and Loss
         2.15. Contacting Web Servers with the HTTP Command
     (3) TERMINAL CONNECTION
         3.1. CONNECT Command Switches
         3.2. Triggers
         3.3. Transparent Printing
         3.4. Binary and Text Session Logs
     (4) FILE TRANSFER AND MANAGEMENT
         4.0. Bug Fixes, Minor Changes, and Clarifications
         4.1. File-Transfer Filename Templates
         4.1.1. Templates in the As-Name
         4.1.2. Templates on the Command Line
         4.1.3. Post-Transfer Renaming
         4.2. File-Transfer Pipes and Filters
         4.2.1. Introduction
         4.2.1.1. Terminology
         4.2.1.2. Notation
         4.2.1.3. Security
         4.2.2. Commands for Transferring from and to Pipes
         4.2.2.1. Sending from a Command
         4.2.2.2. Receiving to a Command
         4.2.3. Using File-Transfer Filters
         4.2.3.1. The SEND Filter
         4.2.3.2. The RECEIVE Filter
         4.2.4. Implicit Use of Pipes
         4.2.5. Success and Failure of Piped Commands
         4.2.6. Cautions about Using Pipes to Transfer Directory Trees
         4.2.7. Pipes and Encryption
         4.2.8. Commands and Functions Related to Pipes
         4.2.8.1. The OPEN !READ and OPEN !WRITE Commands
         4.2.8.2. The REDIRECT Command
         4.2.8.3. Receiving Mail and Print Jobs
         4.2.8.4. Pipe-Related Functions
         4.3. Automatic Per-File Text/Binary Mode Switching
         4.3.1. Exceptions
         4.3.2. Overview
         4.3.3. Commands
         4.3.4. Examples
         4.4. File Permissions
         4.4.1. When ATTRIBUTES PROTECTION is OFF
         4.4.1.1. Unix
         4.4.1.2. VMS
         4.4.2. When ATTRIBUTES PROTECTION is ON
         4.4.2.1. System-Specific Permissions
         4.4.2.1.1. UNIX
         4.4.2.1.2. VMS
         4.4.2.2. System-Independent Permissions
         4.5. File Management Commands
         4.5.1. The DIRECTORY Command
         4.5.2. The CD and BACK Commands
         4.5.2.1. Parsing Improvements
         4.5.2.2. The CDPATH
         4.5.3. Creating and Removing Directories
         4.5.4. The DELETE and PURGE Commands
         4.6. Starting the Remote Kermit Server Automatically
         4.7. File-Transfer Command Switches
         4.7.1. SEND Command Switches
         4.7.2. GET Command Switches
         4.7.3. RECEIVE Command Switches
         4.8. Minor Kermit Protocol Improvements
         4.8.1. Multiple Attribute Packets
         4.8.2. Very Short Packets
         4.9. Wildcard / File Group Expansion
         4.9.1. In UNIX C-Kermit
         4.9.2. In Kermit 95
         4.9.3. In VMS, AOS/VS, OS-9, VOS, etc.
         4.10. Additional Pathname Controls
         4.11. Recursive SEND and GET: Transferring Directory Trees
         4.11.1. Command-Line Options
         4.11.2. The SEND /RECURSIVE Command
         4.11.3. The GET /RECURSIVE Command
         4.11.4. New and Changed File Functions
         4.11.5. Moving Directory Trees Between Like Systems
         4.11.6. Moving Directory Trees Between Unlike Systems
         4.12. Where Did My File Go?
         4.13. File Output Buffer Control
         4.14. Improved Responsiveness
         4.15. Doubling and Ignoring Characters for Transparency
         4.16. New File-Transfer Display Formats
         4.17. New Transaction Log Formats
         4.17.1. The BRIEF Format
         4.17.2. The FTP Format
         4.18. Unprefixing NUL
         4.19. Clear-Channel Protocol
         4.20. Streaming Protocol
         4.20.1. Commands for Streaming
         4.20.2. Examples of Streaming
         4.20.2.1. Streaming on Socket-to-Socket Connections
         4.20.2.2. Streaming on Telnet Connections
         4.20.2.3. Streaming with Limited Packet Length
         4.20.2.4. Streaming on Dialup Connections
         4.20.2.5. Streaming on X.25 Connections
         4.20.3. Streaming - Preliminary Conclusions
         4.21. The TRANSMIT Command
         4.22. Coping with Faulty Kermit Implementations
         4.22.1. Failure to Accept Modern Negotiation Strings
         4.22.2. Failure to Negotiate 8th-bit Prefixing
         4.22.3. Corrupt Files
         4.22.4. Spurious Cancellations
         4.22.5. Spurious Refusals
         4.22.6. Failures during the Data Transfer Phase
         4.22.7. Fractured Filenames
         4.22.8. Bad File Dates
         4.23. File Transfer Recovery
         4.24. FILE COLLISION UPDATE Clarification
         4.25. Autodownload Improvements
     (5) CLIENT / SERVER
         5.0. Hints
         5.1. New Command-Line Options
         5.2. New Client Commands
         5.3. New Server Capabilities
         5.3.1. Creating and Removing Directories
         5.3.2. Directory Listings
         5.4. Syntax for Remote Filenames with Embedded Spaces
         5.5. Automatic Orientation Messages upon Directory Change
         5.6. New Server Controls
         5.7. Timeouts during REMOTE HOST Command Execution
     (6) INTERNATIONAL CHARACTER SETS
         6.0. ISO 8859-15 Latin Alphabet 9
         6.1. The HP-Roman8 Character Set
         6.2. Greek Character Sets
         6.3. Additional Latin-2 Character Sets
         6.4. Additional Cyrillic Character Sets
         6.5. Automatic Character-Set Switching
         6.6. Unicode
         6.6.1. Overview of Unicode
         6.6.2. UCS Byte Order
         6.6.2. UCS Transformation Formats
         6.6.3. Conformance Levels
         6.6.4. Relationship of Unicode with Kermit's Other Character Sets
         6.6.5. Kermit's Unicode Features
         6.6.5.1. File Transfer
         6.6.5.2. The TRANSLATE Command
         6.6.5.3. Terminal Connection
         6.6.5.4. The TRANSMIT Command
         6.6.5.5. Summary of Kermit Unicode Commands
         6.7. Client / Server Character-Set Switching
     (7) SCRIPT PROGRAMMING
         7.0. Bug Fixes
         7.1. The INPUT Command
         7.1.1. INPUT Timeouts
         7.1.2. New INPUT Controls
         7.1.3. INPUT with Pattern Matching
         7.1.4. The INPUT Match Result
         7.2. New or Improved Built-In Variables
         7.3. New or Improved Built-In Functions
         7.4. New IF Conditions
         7.5. Using More than Ten Macro Arguments
         7.6. Clarification of Function Call Syntax
         7.7. Autodownload during INPUT Command Execution
         7.8. Built-in Help for Functions.
         7.9. Variable Assignments
         7.9.1. Assignment Operators
         7.9.2. New Assignment Commands
         7.10. Arrays
         7.10.1. Array Initializers
         7.10.2. Turning a String into an Array of Words
         7.10.3. Arrays of Filenames
         7.10.4. Automatic Arrays
         7.10.5. Sorting Arrays
         7.10.6. Displaying Arrays
         7.10.7. Other Array Operations
         7.10.8. Hints for Using Arrays
         7.10.9. Do-It-Yourself Arrays
         7.10.10. Associative Arrays
         7.11. OUTPUT Command Improvements
         7.12. Function and Variable Diagnostics
         7.13. Return Value of Macros
         7.14. The ASSERT, FAIL, and SUCCEED Commands.
         7.15. Using Alarms
         7.16. Passing Arguments to Command Files
         7.17. Dialogs with Timed Responses
         7.18. Increased Flexibility of SWITCH Case Labels
         7.19. "Kerbang" Scripts
         7.20. IF and XIF Statement Syntax
         7.20.1. The IF/XIF Distinction
         7.20.2. Boolean Expressions (The IF/WHILE Condition)
         7.21. Screen Formatting and Cursor Control
         7.22. Evaluating Arithmetic Expressions
         7.23. Floating-Point Arithmetic
         7.24. Tracing Script Execution
         7.25. Compact Substring Notation
         7.26. New WAIT Command Options
         7.26.1. Waiting for Modem Signals
         7.26.2. Waiting for File Events
         7.27. Relaxed FOR and SWITCH Syntax
     (8) USING OTHER FILE TRANSFER PROTOCOLS
     (9) COMMAND-LINE OPTIONS
         9.0. Extended-Format Command-Line Options
         9.1. Command Line Personalities
         9.2. Built-in Help for Command Line Options
         9.3. New Command-Line Options
    (10) C-KERMIT AND G-KERMIT

III. APPENDICES

III.1. Character Set Tables
III.1.1. The Hewlett Packard Roman8 Character Set
III.1.2. Greek Character Sets
III.1.2.1. The ISO 8859-7 Latin / Greek Alphabet
III.1.2.2. The ELOT 927 Character Set
III.1.2.3. PC Code Page 869
III.2. Updated Country Codes

IV. ERRATA & CORRIGENDA: Corrections to "Using C-Kermit" 2nd Edition.
V. ADDITIONAL COPYRIGHT NOTICES

I. C-KERMIT DOCUMENTATION

   The user manual for C-Kermit is:

     Frank da Cruz and Christine M. Gianone, Using C-Kermit, Second
     Edition, Digital Press / Butterworth-Heinemann, Woburn, MA, 1997,
     622 pages, ISBN 1-55558-164-1.

   The present document is a supplement to Using C-Kermit 2nd Ed, not a
   replacement for it.

   US single-copy price: $52.95; quantity discounts available. Available
   in bookstores or directly from Columbia University:

  The Kermit Project
  Columbia University
  612 West 115th Street
  New York NY  10025-7799
  USA
  Telephone: +1 (212) 854-3703
  Fax:       +1 (212) 662-6442

   Domestic and overseas orders accepted. Price: US $44.95 (US, Canada,
   and Mexico). Shipping: $4.00 within the USA; $15.00 to all other
   countries. Orders may be paid by MasterCard or Visa, or prepaid by
   check in US dollars. Add $65 bank fee for checks not drawn on a US
   bank. Do not include sales tax. Inquire about quantity discounts.

   You can also order by phone from the publisher, Digital Press /
  Butterworth-Heinemann, with MasterCard, Visa, or American Express:

  +1 800 366-2665   (Woburn, Massachusetts office for USA & Canada)
  +44 1865 314627   (Oxford, England distribution centre for UK & Europe)
  +61 03 9245 7111  (Melbourne, Vic, office for Australia & NZ)
  +65 356-1968      (Singapore office for Asia)
  +27 (31) 2683111  (Durban office for South Africa)

   AGerman-language edition of the First Edition is also available:

     Frank da Cruz and Christine M. Gianone, C-Kermit - Einführung und
     Referenz, Verlag Heinz Heise, Hannover, Germany (1994). ISBN
     3-88229-023-4. Deutsch von Gisbert W. Selke. Price: DM 88,00. Verlag
     Heinz Heise GmbH & Co. KG, Helstorfer Strasse 7, D-30625 Hannover.
     Tel. +49 (05 11) 53 52-0, Fax. +49 (05 11) 53 52-1 29.

   TheKermit file transfer protocol is specified in:

     Frank da Cruz, Kermit, A File Transfer Protocol, Digital Press,
     Bedford, MA, 1987, 379 pages, ISBN 0-932376-88-6. US single-copy
     price: $39.95. Availability as above.

   News and articles about Kermit software and protocol are published
   periodically in the journal,Kermit News. Subscriptions are free;
   contact Columbia University at the address above.

   Online news about Kermit is published in the
  comp.protocols.kermit.announce andcomp.protocols.kermit.misc
   newsgroups.

II. NEW FEATURES

   Support for the Bell Labs Plan 9 operating system was added to version
   6.0 too late to be mentioned in the book (although it does appear on
   the cover).

   Specific changes and additions are grouped together by major topic,
   roughly corresponding to the chapters ofUsing C-Kermit.

0. INCOMPATIBILITIES WITH PREVIOUS RELEASES

    1. C-Kermit 7.0 uses FAST Kermit protocol settings by default. This
       includes "unprefixing" of certain control characters. Because of
       this, file transfers that worked with previous releases might not
       work in the new release (but it is more likely that they will work,
       and much faster). If a transfer fails, you'll get a
       context-sensitive hint suggesting possible causes and cures.
       Usually SET PREFIXING ALL does the trick.
    2. C-Kermit 7.0 transfers files in BINARY mode by default. To restore
       the previous behavior, put SET FILE TYPE TEXT in your C-Kermit
       initialization file.
    3. No matter whether FILE TYPE is BINARY or TEXT by default, C-Kermit
       7.0 now switches between text and binary mode automatically on a
       per-file basis according to various criteria, including (a) which
       kind of platform is on the other end of the connection (if known),
       (b) the version of Kermit on the other end, and (c) the file's name
       (see Section 4, especially 4.3). To disable this
       automatic switching and restore the earlier behavior, put SET
       TRANSFER MODE MANUAL in your C-Kermit initialization file. To
       disable automatic switching for a particular transfer, include a
       /TEXT or /BINARY switch with your SEND or GET command.
    4. The RESEND and REGET commands automatically switch to binary mode;
       previously if RESEND or REGET were attempted when FILE TYPE was
       TEXT, these commands would fail immediately, with a message telling
       you they work only when the FILE TYPE is BINARY. Now they simply do
       this for you. See Section 4.23 for additional (important)
       information.
    5. SET PREFIXING CAUTIOUS and MINIMAL now both prefix linefeed (10 and
       138) in case rlogin, ssh, or cu are "in the middle", since
       otherwise <LF>~ might appear in Kermit packets, and this would
       cause rlogin, ssh, or cu to disconnect, suspend, escape back, or
       otherwise wreck the file transfer. Xon and Xoff are now always
       prefixed too, even when Xon/Xoff flow control is not in effect,
       since unprefixing them has proven dangerous on TCP/IP connections.
    6. In UNIX, VMS, Windows, and OS/2, the DIRECTORY command is built
       into C-Kermit itself rather than implemented by running an external
       command or program. The built-in command might not behave the way
       the platform-specific external one did, but many options are
       available for customization. Of course the underlying
       platform-specific command can still be accessed with "!", "@", or
       "RUN" wherever the installation does not forbid. In UNIX, the "ls"
       command can be accessed directly as "ls" in C-Kermit. See
       Section 4.5.1 for details.
    7. SEND ? prints a list of switches rather than a list of filenames.
       If you want to see a list of filenames, use a (system-dependent)
       construction such as SEND ./? (for UNIX, Windows, or OS/2), SEND
       []? (VMS), etc. See Sections 1.5 and4.7.1.
    8. In UNIX, OS-9, and Kermit 95, the wildcard characters in previous
       versions were * and ?. In C-Kermit 7.0 they are *, ?, [, ], {, and
       }, with dash used inside []'s to denote ranges and comma used
       inside {} to separate list elements. If you need to include any of
       these characters literally in a filename, precede each one with
       backslash (\). See Section 4.9.
    9. SET QUIET { ON, OFF } is now on the command stack, just like SET
       INPUT CASE, SET COUNT, SET MACRO ERROR, etc, as described on p.458
       ofUsing C-Kermit, 2nd Edition. This allows any macro or
       command file to SET QUIET ON or OFF without worrying about saving
       and restoring the global QUIET value. For example, this lets you
       write a script that tries SET LINE on lots of devices until it
       finds one free without spewing out loads of error messages, and
       also without disturbing the global QUIET setting, whatever it was.
   10. Because of the new "." operator (which introduces assignments),
       macros whose names begin with "." can not be invoked "by name".
       However, they still can be invoked with DO.
   11. The syntax of the EVALUATE command has changed. SeeSection
       7.9.2. To restore the previous syntax, use SET EVALUATE OLD.
   12. The \v(directory) variable now includes the trailing directory
       separator; in previous releases it did not. This is to allow
       constructions such as:
  cd \v(dir)data.tmp

       to work across platforms that might have different directory
       notation, such as UNIX, Windows, and VMS.
   13. Prior to C-Kermit 7.0, the FLOW-CONTROL setting was global and
       sticky. In C-Kermit 7.0, there is an array of default flow-control
       values for each kind of connection, that are applied automatically
       at SET LINE/PORT/HOST time. Thus a SET FLOW command given before
       SET LINE/PORT/HOST is likely to be undone. Therefore SET FLOW can
       be guaranteed to have the desired effect only if given after the
       SET LINE/PORT/HOST command.
   14. Character-set translation works differently in the TRANSMIT command
       when (a) the file character-set is not the same as the local end of
       the terminal character-set, or (b) when the terminal character-set
       is TRANSPARENT.

1. PROGRAM AND FILE MANAGEMENT AND COMMANDS

1.0. Bug Fixes

   The following patches were issued to correct bugs in C-Kermit 6.0.
   These are described in detail in the 6.0 PATCHES file. All of these
   fixes have been incorporated in C-Kermit 6.1 (never released except as
   K95 1.1.16-17) and 7.0.

 0001   All UNIX         C-Kermit mishandles timestamps on files before 1970
 0002   Solaris 2.5++    Compilation error on Solaris 2.5 with Pro C
 0003   All VMS          CKERMIT.INI Fix for VMS
 0004   VMS/VAX/UCX 2.0  C-Kermit 6.0 can't TELNET on VAX/VMS with UCX 2.0
 0005   All              C-Kermit Might Send Packets Outside Window
 0006   All              MOVE from SEND-LIST does not delete original files
 0007   Solaris 2.5++    Higher serial speeds on Solaris 2.5
 0008   All              C-Kermit application file name can't contain spaces
 0009   AT&T 7300 UNIXPC setuid and hardware flow-control problems
 0010   Linux on Alpha   Patch to make ckutio.c compile on Linux/Alpha
 0011   OS-9/68000 2.4   Patch to make ck9con.c compile on OS-9/68000 2.4
 0012   MW Coherent 4.2  Patches for successful build on Coherent 4.2
 0013   SINIX-Y 5.43     "delay" variable conflicts with <sys/clock.h>
 0014   VMS/VAX/CMU-IP   Subject: Patches for VAX/VMS 5.x + CMU-IP
 0015   All              XECHO doesn't flush its output
 0016   VMS              CD and other directory operations might not work
 0017   Linux 1.2.x++    Use standard POSIX interface for high serial speeds
 0018   UNIX             SET WILDCARD-EXPANSION SHELL dumps core
 0019   All              Hayes V.34 modem init string problem
 0020   All              READ command does not fail if file not open
 0021   All              Problems with long function arguments
 0022   All              Certain \function()s can misbehave
 0023   All              X MOD 0 crashes program
 0024   All              Internal bulletproofing for lower() function
 0025   OpenBSD          Real OpenBSD support for C-Kermit 6.0
 0026   All              Incorrect checks for macro/command-file nesting depth
 0027   All              ANSWER doesn't automatically CONNECT
 0028   All              Overzealous EXIT warning
 0029   All              OUTPUT doesn't echo when DUPLEX is HALF
 0030   All              Minor problems with REMOTE DIRECTORY/DELETE/etc
 0031   All              CHECK command broken
 0032   All              Problem with SET TRANSMIT ECHO
 0033   UNIX, VMS, etc   HELP SET SERVER says too much
 0034   All              READ and !READ too picky about line terminators
 0035   All              END from inside SWITCH doesn't work
 0036   All              Problem telnetting to multihomed hosts
 0037   All              Redirection failures in REMOTE xxx > file

   REDIRECT was missing in many UNIX C-Kermit implementations; in version
   7.0, it should be available in all of them.

1.1. Command Continuation

   Comments that start with ";" or "#" can no longer be continued. In:

  ; this is a comment -
  echo blah

   the ECHO command will execute, rather than being taken as a
   continuation of the preceding comment line. This allows easy
   "commenting out" of commands from macro definitions.

   However, the text of the COMMENT command can still be continued onto
   subsequent lines:

  comment this is a comment -
  echo blah

   As of version 6.0, backslash is no longer a valid continuation
   character. Only hyphen should be used for command continuation. This is
   to make it possible to issue commands like "cd a:\" on DOS-like
   systems.

   As of version 7.0:

     * You can quote a final dash to prevent it from being a continuation
       character:
  echo foo\-

       This prints "foo-". The command is not continued.
     * You can enter commands such as:
  echo foo - ; this is a comment

       interactively and they are properly treated as continued commands.
       Previously this worked only in command files.

1.2. Editor Interface

   SET EDITOR name [ options ]
          Lets you specify a text-editing program. The name can be a fully
          specified pathname like /usr/local/bin/emacs19/emacs, or it can
          be the name of any program in your PATH, e.g. "set editor
          emacs". In VMS, it must be a DCL command like "edit",
          "edit/tpu", "emacs", etc. If an environment variable EDITOR is
          defined when Kermit starts, its value is the default editor. You
          can also specify options to be included on the editor command
          line. Returns to Kermit when the editor exits.

   EDIT [ filename ]
          If the EDIT command is given without a filename, then if a
          previous filename had been given to an EDIT command, it is used;
          if not, the editor is started without a file. If a filename is
          given, the editor is started on that file, and the filename is
          remembered for subsequent EDIT commands.

   SHOW EDITOR
          Displays the full pathname of your text editor, if any, along
          with any command line options, and the file most recently edited
          (and therefore the default filename for your next EDIT command).

   Related variables: \v(editor), \v(editopts), \v(editfile).

1.3. Web Browser and FTP Interface

   C-Kermit includes an FTP command, which simply runs the FTP program;
   C-Kermit does not include any built-in support for Internet File
   Transfer Protocol, nor any method for interacting directly with an FTP
   server. In version 7.0, however, C-Kermit lets you specify your FTP
   client:

   SET FTP-CLIENT [ name [ options ] ]
          The name is the name of the FTP executable. In UNIX, Windows, or
          OS/2, it can be the filename of any executable program in your
          PATH (e.g. "ftp.exe" in Windows, "ftp" in UNIX); elsewhere (or
          if you do not have a PATH definition), it must be the fully
          specified pathname of the FTP program. If the name contains any
          spaces, enclose it braces. Include any options after the
          filename; these depend the particular ftp client.

   The Web browser interface is covered in the following subsections.

1.3.1. Invoking your Browser from C-Kermit

   BROWSE [ url ]
          Starts your preferred Web browser on the URL, if one is given,
          otherwise on the most recently given URL, if any. Returns to
          Kermit when the browser exits.

   SET BROWSER [ name [ options ] ]
          Use this command to specify the name of your Web browser
          program, for example: "set browser lynx". The name must be in
          your PATH, or else it must be a fully specified filename; in VMS
          it must be a DCL command.

   SHOW BROWSER
          Displays the current browser, options, and most recent URL.

   Related variables: \v(browser), \v(browsopts), \v(browsurl).

   Also see Section 2.15: Contacting Web Servers with the HTTP
   Command.

1.3.2. Invoking C-Kermit from your Browser

   The method for doing this depends, of course, on your browser. Here are
   some examples:

   Netscape on UNIX (X-based)
          In the Options->Applications section, set your Telnet
          application to:

  xterm -e /usr/local/bin/kermit/kermit -J %h %p

          (replace "/usr/local/bin/kermit/kermit" by C-Kermit's actual
          pathname). -J is C-Kermit's command-line option to "be like
          Telnet"; %h and %p are Netscape placeholders for hostname and
          port.

   Lynx on UNIX
          As far as we know, this can be done only at compile time. Add
          the following line to the Lynx userdefs.h file before building
          the Lynx binary:

  #define TELNET_COMMAND "/opt/bin/kermit -J"

          And then add lines like the following to the Lynx.cfg file:

  DOWNLOADER:Kermit binary download:/opt/bin/kermit -i -V -s %s -a %s:TRUE
  DOWNLOADER:Kermit text download:/opt/bin/kermit -s %s -a %s:TRUE

  UPLOADER:Kermit binary upload:/opt/bin/kermit -i -r -a %s:TRUE
  UPLOADER:Kermit text upload:/opt/bin/kermit -r -a %s:TRUE
  UPLOADER:Kermit text get:/opt/bin/kermit -g %s:TRUE
  UPLOADER:Kermit binary get:/opt/bin/kermit -ig %s:TRUE

   But none of the above is necessary if you make C-Kermit your default
   Telnet client, which you can do by making a symlink called 'telnet' to
   the C-Kermit 7.0 binary. See Section 9.1 for details.

1.4. Command Editing

   Ctrl-W ("Word delete") was changed in 7.0 to delete back to the
   previous non-alphanumeric, rather than all the way back to the previous
   space.

1.5. Command Switches

   As of version 7.0, C-Kermit's command parser supports a new type of
   field, called a "switch". This is an optional command modifier.

1.5.1. General Switch Syntax

   A switch is a keyword beginning with a slash (/). If it takes a value,
   then the value is appended to it (with no intervening spaces),
   separated by a colon (:) or equal sign (=). Depending on the switch,
   the value may be a number, a keyword, a filename, a date/time, etc.
   Examples:

  send oofa.txt                              ; No switches
  send /binary oofa.zip                      ; A switch without a value
  send /protocol:zmodem oofa.zip             ; A switch with a value (:)
  send /protocol=zmodem oofa.zip             ; A switch with a value (=)
  send /text /delete /as-name:x.x oofa.txt   ; Several switches

   Like other command fields, switches are separated from other fields,
   and from each other, by whitespace, as shown in the examples just
   above. You can not put them together like so:

  send/text/delete/as-name:x.x oofa.txt

   (as you might do in VMS or DOS, or as we might once have done in
   TOPS-10 or TOPS0-20, or PIP). This is primarily due to ambiguity
   between "/" as switch introducer versus "/" as UNIX directory
   separator; e.g. in:

  send /delete/as-name:foo/text oofa.txt

   Does "foo/text" mean the filename is "foo" and the transfer is to be in
   text mode, or does it mean the filename is "foo/text"? Therefore we
   require whitespace between switches to resolve the ambiguity. (That's
   only one of several possible ambiguities -- it is also conceivable that
   a file called "text" exists in the path "/delete/as-name:foo/").

   In general, if a switch can take a value, but you omit it, then either
   a reasonable default value is supplied, or an error message is printed:

  send /print:-Plaserwriter oofa.txt         ; Value included = print options
  send /print oofa.txt                       ; Value omitted, OK
  send /mail:kermit@columbia.edu oofa.txt    ; Value included = address
  send /mail oofa.txt                        ; Not OK - address required
  ?Address required

   Context-sensitive help (?) and completion (Esc or Tab) are available in
   the normal manner:

  C-Kermit> send /pr? Switch, one of the following:
    /print /protocol
  C-Kermit> send /pro<ESC>tocol:?  File-transfer protocol,
   one of the following:
    kermit   xmodem   ymodem   ymodem-g   zmodem
  C-Kermit> send /protocol:k<TAB>ermit

   If a switch takes a value and you use completion on it, a colon (:) is
   printed at the end of its name to indicate this. If it does not take a
   value, a space is printed.

   Also, if you type ? in a switch field, switches that take values are
   shown with a trailing colon; those that don't take values are shown
   without one.

1.5.2. Order and Effect of Switches

   The order of switches should not matter, except that they are evaluated
   from left to right, so if you give two switches with opposite effects,
   the rightmost one is used:

  send /text /binary oofa.zip                ; Sends oofa.zip in binary mode.

   Like other command fields, switches have no effect whatsoever until the
   command is entered (by pressing the Return or Enter key). Even then,
   switches affect only the command with which they are included; they do
   not have global effect or side effects.

1.5.3. Distinguishing Switches from Other Fields

   All switches are optional. A command that uses switches lets you give
   any number of them, including none at all. Example:

  send /binary oofa.zip
  send /bin /delete oofa.zip
  send /bin /as-name:mupeen.zip oofa.zip
  send oofa.zip

   But how does Kermit know when the first "non-switch" is given? It has
   been told to look for both a switch and for something else, the data
   type of the next field (filename, number, etc). In most cases, this
   works well. But conflicts are not impossible. Suppose, for example, in
   UNIX there was a file named "text" in the top-level directory. The
   command to send it would be:

  send /text

   But C-Kermit would think this was the "/text" switch. To resolve the
   conflict, use braces:

  send {/text}

   or other circumlocutions such as "send //text", "send /./text", etc.

   The opposite problem can occur if you give an illegal switch that
   happens to match a directory name. For example:

  send /f oofa.txt

   There is no "/f" switch (there are several switches that begin with
   "/f", so "/f" is ambiguous). Now suppose there is an "f" directory in
   the root directory; then this command would be interpreted as:

     Send all the files in the "/f" directory, giving each one an as-name
     of "oofa.txt".

   This could be a mistake, or it could be exactly what you intended;
   C-Kermit has no way of telling the difference. To avoid situations like
   this, spell switches out in full until you are comfortable enough with
   them to know the minimum abbreviation for each one. Hint: use ? and
   completion while typing switches to obtain the necessary feedback.

1.5.4. Standard File Selection Switches

   The following switches are used on different file-oriented commands
   (such as SEND, DIRECTORY, DELETE, PURGE) to refine the selection of
   files that match the given specification.

   /AFTER:date-time
          Select only those files having a date-time later than the one
          given. See Section 1.6 for date-time formats. Synonym:
          /SINCE.

   /NOT-AFTER:date-time
          Select only those files having a date-time not later than (i.e.
          earlier or equal to) the one given. Synonym: /NOT-SINCE.

   /BEFORE:date-time
          Select only those files having a date-time earlier than the one
          given.

   /NOT-BEFORE:date-time
          Select only those files having a date-time not earlier than
          (i.e. later or equal to) the one given.

   /DOTFILES
          UNIX and OS-9 only: The filespec is allowed to match files whose
          names start with (dot) period. Normally these files are not
          shown.

   /NODOTFILES
          (UNIX and OS-9 only) Don't show files whose names start with dot
          (period). This is the opposite of /DOTFILES, and is the default.
          Note that when a directory name starts with a period, the
          directory and (in recursive operations) all its subdirectories
          are skipped.

   /LARGER-THAN:number
          Only select files larger than the given number of bytes.

   /SMALLER-THAN:number
          Only select files smaller than the given number of bytes.

   /EXCEPT:pattern
          Specifies that any files whose names match the pattern, which
          can be a regular filename, or may contain "*" and/or "?"
          metacharacters (wildcards), are not to be selected. Example:

  send /except:*.log *.*

          sends all files in the current directory except those with a
          filetype of ".log". Another:

  send /except:*.~*~ *.*

          sends all files except the ones that look like Kermit or EMACS
          backup files (such as "oofa.txt.~17~") (of course you can also
          use the /NOBACKUP switch for this).

          The pattern matcher is the same one used by IF MATCH string
          pattern ( Section 7.4), so you can test your patterns using
          IF MATCH. If you need to match a literal * or ? (etc), precede
          it by a backslash (\). If the pattern contains any spaces, it
          must be enclosed in braces:

  send /except:{Foo bar} *.*

          The pattern can also be a list of up to 8 patterns. In this
          case, the entire pattern must be enclosed in braces, and each
          sub-pattern must also be enclosed in braces; this eliminates the
          need for designating a separator character, which is likely to
          also be a legal filename character on some platform or other,
          and therefore a source of confusion. You may include spaces
          between the subpatterns but they are not necessary. The
          following two commands are equivalent:

  send /except:{{ck*.o} {ck*.c}} ck*.?
  send /except:{{ck*.o}{ck*.c}} ck*.?

          If a pattern is to include a literal brace character, precede it
          with "\". Also note the apparent conflict of this list format
          and the string-list format described in Section 4.9.1. In
          case you want to include a wildcard string-list with braces on
          its outer ends as an /EXCEPT: argument, do it like this:

  send /except:{{{ckuusr.c,ckuus2.c,ckuus6.c}}} ckuus*.c

1.5.5. Setting Preferences for Different Commands

   Certain oft-used commands offer lots of switches because different
   people have different requirements or preferences. For example, some
   people want to be able to delete files without having to watch a list
   of the deleted files scroll past, while others want to be prompted for
   permission to delete each file. Different people prefer different
   directory-listing styles. And so on. Such commands can be tailored with
   the SET OPTIONS command:

   SET OPTIONS command [ switch [ switch [ ... ] ] ]
          Sets each switch as the default for the given command, replacing
          the "factory default". Of course you can also override any
          defaults established by the SET OPTIONS command by including the
          relevant switches in the affected command any time you issue it.

   SHOW OPTIONS
          Lists the commands that allows option-setting, and the options
          currently in effect, if any, for each. Switches that have
          synonyms are shown under their primary name; for example. /LOG
          and /VERBOSE are shown as /LIST.

   Commands for which options may be set include DIRECTORY, DELETE, PURGE,
   and TYPE. Examples:

  SET OPTIONS DIRECTORY /PAGE /NOBACKUP /HEADING /SORT:DATE /REVERSE
  SET OPTIONS DELETE /LIST /NOHEADING /NOPAGE /NOASK /NODOTFILES
  SET OPTIONS TYPE /PAGE

   Not necessarily all of a command's switches can be set as options. For
   example, file selection switches, since these would normally be
   different for each command.

   Put the desired SET OPTIONS commands in your C-Kermit customization
   file for each command whose default switches you want to change every
   time you run C-Kermit.

1.6. Dates and Times

   Some commands and switches take date-time values, such as:

  send /after:{8-Feb-2000 10:28:01}

   Various date-time formats are acceptable. The rules for the date are:

     * The year must have 4 digits.
     * If the year comes first, the second field is the month.
     * The day, month, and year may be separated by spaces, /, -, or
       underscore.
     * The month may be numeric (1 = January) or spelled out or
       abbreviated in English.

   If the date-time string contains any spaces, it must be enclosed in
   braces. Examples of legal dates:

                           Interpretation:
  2000-Feb-8                8 February 2000
  {2000 Feb 8}              8 February 2000
  2000/Feb/8                8 February 2000
  2000_Feb_8                8 February 2000
  2000-2-8                  8 February 2000
  2000-02-08                8 February 2000
  8-Feb-2000                8 February 2000
  08-Feb-2000               8 February 2000
  12/25/2000                25 December 2000
  25/12/2000                25 December 2000

   The last two examples show that when the year comes last, and the month
   is given numerically, the order of the day and month doesn't matter as
   long as the day is 13 or greater (mm/dd/yyyy is commonly used in the
   USA, whereas dd/mm/yyyy is the norm in Europe). However:

  08/02/2000                Is ambiguous and therefore not accepted.

   If a date is given, the time is optional and defaults to 00:00:00. If
   the time is given with a date, it must follow the date, separated by
   space, /, -, or underscore, and with hours, minutes, and seconds
   separated by colon (:). Example:

  2000-Feb-8 10:28:01       Represents 8 February 2000, 10:28:01am

   If a date is not given, the current date is used and a time is
   required.

   Time format is hh:mm:ss or hh:mm or hh in 24-hour format, or followed
   by "am" or "pm" (or "AM" or "PM") to indicate morning or afternoon.
   Examples of times that are acceptable:

                           Interpretation:
  3:23:56                    3:23:56am
  3:23:56am                  3:23:56am
  3:23:56pm                  3:23:56pm = 15:23:56
 15:23:56                    3:23:56pm = 15:23:56
  3:23pm                     3:23:00pm = 15:23:00
  3:23PM                     3:23:00pm = 15:23:00
  3pm                        3:00:00pm = 15:00:00

   Examples of legal date-times:

  send /after:{8 Feb 2000 10:28:01}
  send /after:8_Feb_2000_10:28:01
  send /after:8-Feb-2000/10:28:01
  send /after:2000/02/08/10:28:01
  send /after:2000/02/08_10:28:01
  send /after:2000/02/08_10:28:01am
  send /after:2000/02/08_10:28:01pm
  send /after:2000/02/08_10:28pm
  send /after:2000/02/08_10pm
  send /after:10:00:00pm
  send /after:10:00pm
  send /after:10pm
  send /after:22

   Finally, there is a special all-numeric format you can use:

  yyyymmdd hh:mm:ss

   For example:

  20000208 10:28:01

   This is Kermit's standard date-time format (based on ISO 8601), and is
   accepted (among other formats) by any command or switch that requires a
   date-time, and is output by any function whose result is a calendar
   date-time.

   There are no optional parts to this format and it must be exactly 17
   characters long, punctuated as shown (except you can substitute
   underscore for space in contexts where a single "word" is required).
   The time is in 24-hour format (23:00:00 is 11:00pm). This is the format
   returned by \fdate(filename), so you can also use constructions like
   this:

  send /after:\fdate(oofa.txt)

   which means "all files newer than oofa.txt".

   Besides explicit dates, you can also use the any of the following
   shortcuts:

   TODAY
          Stands for the current date at 00:00:00.

   TODAY 12:34:56
          Stands for the current date at the given time.

   YESTERDAY
          Stands for yesterday's date at 00:00:00. A time may also be
          given.

   TOMORROW
          Stands for tomorrow's date at 00:00:00. A time may also be
          given.

   + number { DAYS, WEEKS, MONTHS, YEARS } [ time ]
          Is replaced by the future date indicated, relative to the
          current date. If the time is omitted, 00:00:00 is used.
          Examples: +3days, +2weeks, +1year, +37months.

   - number { DAYS, WEEKS, MONTHS, YEARS } [ time ]

          Is replaced by the past date indicated, relative to the current
          date. If the time is omitted, 00:00:00 is used.

   The time can be separated from the date shortcut by any of the same
   separators that are allowed for explicit date-times: space, hyphen,
   slash, period, or underscore. In switches and other space-delimited
   fields, use non-spaces to separate date/time fields, or enclose the
   date-time in braces, e.g.:

  purge /before:-4days_12:00:00
  purge /before:{- 4 days 12:00:00}

   Of course you can also use variables:

  define \%n 43
  purge /before:-\%ndays_12:00:00

   Shortcut names can be abbreviated to any length that still
   distinguishes them from any other name that can appear in the same
   context, e.g. "TOD" for today, "Y" for yesterday. Also, the special
   abbreviation "wks" is accepted for WEEKS, and "yrs" for "YEARS".

   (To see how to specify dates relative to a specific date, rather than
   the current one, see the\fmjd() function description below.)

   You can check date formats with the DATE command. DATE by itself prints
   the current date and time in standard format: yyyymmdd hh:mm:ss. DATE
   followed by a date and/or time (including shortcuts) converts it to
   standard format if it can understand it, otherwise it prints an error
   message.

   The following variables and functions deal with dates and times; any
   function argument designated as "date-time" can be in any of the
   formats described above.

   \v(day)
          The first three letters of the English word for the current day
          of the week, e.g. "Wed".

   \fday(date-time)
          The first three letters of the English word for day of the week
          of the given date. If a time is included, it is ignored.
          Example: \fday(8 Feb 1988) = "Mon".

   \v(nday)
          The numeric day of the week: 0 = Sunday, 1 = Monday, ..., 6 =
          Saturday.

   \fnday(date-time)
          The numeric day of the week for the given date. If a time is
          included, it is ignored. Example: \fnday(8 Feb 1988) = "1".

   \v(date)
          The current date as dd mmm yyyy, e.g. "08 Feb 2000" (as in this
          example, a leading zero is supplied for day-of-month less than
          10).

   \v(ndate)
          The current date in numeric format: yyyymmdd, e.g. "20000208".

   \v(time)
          The current time as hh:mm:ss, e.g. "15:27:14".

   \ftime(time)
          The given free-format date and/or time (e.g. "3pm") returns the
          time (without the date) converted to hh:mm:ss 24-hour format,
          e.g. "15:00:00" (the date, if given, is ignored).

   \v(ntime)
          The current time as seconds since midnight, e.g. "55634".

   \v(tftime)
          The elapsed time of the most recent file-transfer operation in
          seconds.

   \v(intime)
          The elapsed time for the most recent INPUT command to complete,
          in milliseconds.

   \fntime(time)
          The given free-format date and/or time is converted to seconds
          since midnight (the date, if given, is ignored). This function
          replaces \ftod2secs(), which is now a synonym for \fntime().
          Unlike \ftod2secs(), \fntime() allows a date to be included, and
          it allows the time to be in free format (like 3pm), and it
          allows the amount of time to be more than 24 hours. E.g.
          \fntime(48:00:00) = 172800. Example of use:

  set alarm \fntime(48:00:00) ; set alarm 48 hours from now.

   \fn2time(seconds)
          The given number of seconds is converted to hh:mm:ss format.

   \fdate(filename)
          Returns the modification date-time of the given file in standard
          format: yyyymmdd hh:mm:ss.

   \fcvtdate(date-time)
          Converts a free-format date and/or time to Kermit standard
          format: yyyymmdd hh:mm:ss. If no argument is given, returns the
          current date-time in standard format. If a date is given but no
          time, the converted date is returned without a time. If a time
          is given with no date, the current date is supplied. Examples:

  \fcvtdate(4 Jul 2000 2:21:17pm) = 20000704 14:21:17
  \fcvtdate() = 20000704 14:21:17 (on 4 Jul 2000 at 2:21:17pm).
  \fcvtd(4 Jul 2000) = 20000704
  \fcvtd(6pm) = 20000704 18:00:00 (on 4 Jul 2000 at 6:00pm).

   \fdayofyear(date-time)
   \fdoy(date-time)
          Converts a free-format date and/or time to yyyyddd, where ddd is
          the 3-digit day of the year, and 1 January is Day 1. If a time
          is included with the date, it is returned in standard format. If
          a date is included but no time, the date is returned without a
          time. If a time is given with no date, the time is converted and
          the current date is supplied. If no argument is given, the
          current date-time is returned. Synonym: \fdoy(). Examples:

  \fddayofyear(4 Jul 2000 2:21:17pm) = 2000185 14:21:17
  \fdoy() = 2000185 14:21:17 (on 4 Jul 2000 at 2:21:17pm).
  \fdoy(4 Jul 2000) = 2000185
  \fdoy(6pm) = 2000185 18:00:00 (on 4 Jul 2000 at 6:00pm).

   Note: The yyyyddd day-of-year format is often erroneously referred to
   as a Julian date. However, a true Julian date is a simple counting
   number, the number of days since a certain fixed day in the past.
  See \fmjd() below.

   \fdoy2date(date-time)
          Converts a date or date-time in day-of-year format to a standard
          format date. A yyyyddd-format date must be supplied; time is
          optional. The given date is converted to yyyymmdd format. If a
          time is given, it is converted to 24-hour format. Examples:

  \fdoy2date(2000185) = 20000704
  \fdoy2(2000185 3pm) = 20000704 15:00:00

   \fmjd(date-time)
          Converts free-format date and/or time to a Modified Julian Date
          (MJD), the number of days since 17 Nov 1858 00:00:00. If a time
          is given, it is ignored. Examples:

  \fmjd(4 Jul 2000) = 50998
  \fmjd(17 Nov 1858) = 0
  \fmjd(16 Nov 1858) = -1

   \fmjd2date(mjd)
          Converts an MJD (integer) to standard date format, yyyymmdd:

  \fmjd2(50998) = 4 Jul 1998
  \fmjd2(0) = 17 Nov 1858
  \fmjd2(-1) = 16 Nov 1858
  \fmjd2(-365) = 17 Nov 1857

   MJDs are normal integers and, unlike DOYs, may be added, subtracted,
   etc, with each other or with other integers, to obtain meaningful
   results. For example, to find out the date 212 days ago:

  echo \fmjd2date(\fmjd()-212)

   Constructions such as this can be used in any command where a date-time
   is required, e.g.:

  send /after:\fmjd2date(\fmjd()-212)

   to send all files that are not older than 212 days (this is equivalent
   to "send /after:-212days").

   MJDs also have other regularities not exhibited by other date formats.
   For example, \fmodulus(\fmjd(any-date),7) gives the day of the week for
   any date (where 4=Sun, 5=Mon, ..., 3=Sat). (However, it is easier to
   use \fnday() for this purpose, and it gives the more conventional
   result of 0=Sun, 1=Mon, ..., 6=Sat).

   Note that if MJDs are to be compared, they must be compared numerically
   (IF <, =, >) and not lexically (IF LLT, EQUAL, LGT), whereas DOYs must
   be compared lexically if they include a time (which contains ":"
   characters); however, if DOYs do not include a time, they may also be
   compared numerically.

   In any case, lexical comparison of DOYs always produces the appropriate
   result, as does numeric comparison of MJDs.

   The same comments apply to sorting. Also note that DOYs are fixed
   length, but MJDs can vary in length. However, all MJDs between 3 April
   1886 and 30 Aug 2132 are 5 decimal digits long. (MJDs become 6 digits
   long on 31 Aug 2132, and 7 digits long on 13 Oct 4596).

1.7. Partial Completion of Keywords

   Partial completion of keywords was added in C-Kermit 7.0. In prior
   versions, if completion was attempted (by pressing the Esc or Tab key)
   on a string that matched different keywords, you'd just get a beep. Now
   Kermit completes up to the first character where the possibly matching
   keywords differ and then beeps. For example:

  C-Kermit> send /n<Tab>

   which matches /NOT-BEFORE and /NOT-AFTER, now completes up to the dash:

  C-Kermit> send /n<Tab>ot-<Beep>

   Partial completion works for filenames too (as it has for some years).

1.8. Command Recall

   C-Kermit has had a command history buffer for some time, which could be
   scrolled interactively using control characters or (in Kermit 95 only)
   arrow keys. Version 7.0 adds a REDO command that allows the most recent
   command matching a given pattern to be re-executed:

   { REDO, RR, ^ } [ pattern ]
          Search the command history list for the most recent command that
          matches the given pattern, and if one is found, execute it
          again.

   The pattern can be a simple string (like "send"), in which case the
   last SEND command is re-executed. Or it can contain wildcard characters
   "*" and/or "?", which match any string and any single character,
   respectively (note that "?" must be preceded by backslash to override
   its normal function of giving help), and in most C-Kermit versions may
   also include [] character lists and {} string lists (seeSection
   4.9).

   The match works by appending "*" to the end of the given pattern (if
   you didn't put one there yourself). Thus "redo *oofa" becomes "redo
   *oofa*" and therefore matches the most recent command that contains
   "oofa" anywhere within the command. If you want to inhibit the
   application of the trailing "*", e.g. to force matching a string at the
   end of a command, enclose the pattern in braces:

  redo {*oofa}

   matches the most recent command that ends with "oofa".

   REDO commands themselves are not entered into the command history list.
   If no pattern is given, the previous (non-REDO) command is re-executed.
   The REDOne command is reinserted at the end of the command history
   buffer, so the command scrollback character (Ctrl-P, Ctrl-B, or
   Uparrow) can retrieve it.

   Examples:

  C-Kermit> echo foo
  foo
  C-Kermit> show alarm
  (no alarm set)
  C-Kermit> echo blah
  blah
  C-Kermit> redo          ; Most recent command
  blah
  C-Kermit> redo s        ; Most recent command starting with "s"
  (no alarm set)
  C-Kermit> redo echo f   ; Most recent command starting with "echo f"
  foo
  C-Kermit> redo *foo     ; Most recent command that has "foo" in it
  foo
  C-Kermit> <Ctrl-P>      ; Scroll back
  C-Kermit> echo foo      ; The REDOne command is there
  C-Kermit> redo {*foo}   ; Most recent command that ends with "foo"
  foo
  C-Kermit>

   Since REDO, REDIAL, and REDIRECT all start the same way, and RED is the
   designated non-unique abbreviation for REDIAL, REDO must be spelled out
   in full. For convenience, RR is included as an invisible easy-to-type
   synonym for REDO. You can also use the "^" character for this:

  C-Kermit> ^             ; Most recent command
  C-Kermit> ^ s           ; Most recent command starting with "s"
  C-Kermit> ^s            ; Ditto (space not required after "^").
  C-Kermit> ^*foo         ; Most recent command that has "foo" in it.
  C-Kermit> ^{*foo}       ; Most recent command ends with "foo".

   Unlike the manual command-history-scrolling keys, the REDO command can
   be used in a script, but it's not recommended (since the command to be
   REDOne might not be found, so if the REDO command fails, you can't tell
   whether it was because REDO failed to find the requested command, or
   because the command was found but it failed).

1.9. EXIT Messages

   The EXIT and QUIT commands now accept an optional message to be
   printed. This makes the syntax of EXIT and QUIT just like END and STOP:

   { EXIT, QUIT, END, STOP } [ status-code [ message ] ]

   where status-code is a number (0 indicating success, nonzero indicating
   failure). This is handy in scripts that are never supposed to enter
   interactive mode:

  dial 7654321
  if fail exit 1 Can't make connection - try again later.

   Previously this could only be done in two steps:

  dial 7654321
  xif fail { echo Can't make connection - try again later, exit 1 }

   A status code must be included in order to specify a message. In the
   case of EXIT and QUIT, the default status code is contained in the
   variable \v(exitstatus), and is set automatically by various events
   (file transfer failures, etc; it can also be set explicitly with the
   SET EXIT STATUS command). If you want to give an EXIT or QUIT command
   with a message, but without changing the exit status from what it
   normally would have been, use the \v(exitstatus) variable, e.g.:

   exit \v(exitstatus) Goodbye from \v(cmdfile).

   The EXIT status is returned to the system shell or whatever other
   process invoked C-Kermit, e.g. in UNIX:

  C-Kermit> exit 97 bye bye
  bye bye
  $ echo $?
  97
  $

1.10. Managing Keyboard Interruptions

   When C-Kermit is in command or file-transfer mode (as opposed to
   CONNECT mode), it can be interrupted with Ctrl-C. Version 7.0 adds the
   ability to disarm the Ctrl-C interrupt:

   SET COMMAND INTERRUPT { ON, OFF }
          COMMAND INTERRUPT is ON by default, meaning the Ctrl-C can be
          used to interrupt a command or a file transfer in progress. Use
          OFF to disable these interruptions, and use it with great
          caution for obvious reasons.

   SET TRANSFER INTERRUPT { ON, OFF }
          This can be used to disable keyboard interruption of file
          transfer when C-Kermit is in local mode, or to re-enable it
          after it has been disabled. This applies to the X, Z, E, and
          similar keys as well as to the system interrupt character,
          usually Ctrl-C. This is distinct from SET TRANSFER CANCELLATION,
          which tells whether packet mode can be exited by sending a
          special sequence of characters.

   Several other commands can be interrupted by pressing any key while
   they are active. Version 7.0 adds the ability to disable this form of
   interruption also:

   SET INPUT CANCELLATION { ON, OFF }
          Whether an INPUT command in progress can be interrupted by
          pressing a key. Normally ON. Setting INPUT CANCELLATION OFF
          makes INPUT commands uninterruptible except by Ctrl-C (unless
          COMMAND INTERRUPTION is also OFF).

   SET SLEEP CANCELLATION { ON, OFF }
          Whether a SLEEP, PAUSE, or WAIT command in progress can be
          interrupted by pressing a key. Normally ON. Setting SLEEP
          CANCELLATION OFF makes these commands uninterruptible except by
          Ctrl-C (unless COMMAND INTERRUPTION is also OFF). Synonyms: SET
          PAUSE CANCELLATION, SET WAIT CANCELLATION.

   So to make certain a script is not interruptible by the user, include
   these commands:

  SET TRANSFER INTERRUPT OFF
  SET SLEEP CANCELLATION OFF
  SET INPUT CANCELLATION OFF
  SET COMMAND INTERRUPTION OFF

   Make sure to turn them back on afterwards if interruption is to be
   re-enabled.

   When a PAUSE, SLEEP, WAIT, or INPUT command is interrupted from the
   keyboard, the new variable \v(kbchar) contains a copy of the (first)
   character that was typed and caused the interruption, provided it was
   not the command interrupt character (usually Ctrl-C). If these commands
   complete successfully or time out without a keyboard interruption, the
   \v(kbchar) variable is empty.

   The \v(kbchar) variable (like any other variable) can be tested with:

  if defined \v(kbchar) command

   The command is executed if the variable is not empty.

   The \v(kbchar) variable can be reset with WAIT 0 (PAUSE 0, SLEEP 0,
   etc).

1.11. Taming The Wild Backslash -- Part Deux

  Using C-Kermit, 2nd Edition, contains a brief section, "Taming the
   Wild Backslash", on page 48, which subsequent experience has shown to
   be inadequate for Kermit users intent on writing scripts that deal with
   Windows, DOS, and OS/2 filenames, in which backslash (\) is used as the
   directory separator. This section fills in the blanks.

1.11.1. Background

   The Kermit command language shares a certain unavoidable but annoying
   characteristic with most other command languages that are capable of
   string replacement, namely the necessity to "quote" certain characters
   when you want them to be taken literally. This is a consequence of the
   facts that:

    1. One or more characters must be set aside to denote replacement,
       rather than acting as literal text.
    2. We have only 96 printable characters to work with in ASCII, which
       is still the only universally portable character set.
    3. There is no single printable character that is unused everywhere.
    4. Variables are not restricted to certain contexts, as they are in
       formal programming languages like C and Fortran, but can appear
       anywhere at all within a command, and therefore require special
       syntax.

   Thus there can be conflicts. To illustrate, the standard UNIX shell
   uses dollar sign ($) to introduce variables. So the shell command:

  echo $TERM

   displays the value of the TERM variable, e.g. vt320. But suppose you
   want to display a real dollar sign:

  echo The price is $10.20

   This causes the shell to evaluate the variable "$1", which might or
   might not exist, and substitute its value, e.g.:

  The price is 0.20

   (in this case the $1 variable had no value.) This is probably not what
   you wanted. To force the dollar sign to be taken literally, you must
   apply a "quoting rule", such as "precede a character by backslash (\)
   to force the shell to take the character literally":

  echo The price is \$10.20
  The price is $10.20

   But now suppose you want the backslash AND the dollar sign to be taken
   literally:

  echo The price is \\$10.20

   This doesn't work, since the first backslash quotes the second one,
   thereby leaving the dollar sign unquoted again:

  The price is \0.20

   Quoting the dollar sign requires addition of a third backslash:

  echo The price is \\\$10.20
  The price is \$10.20

   The first backslash quotes the second one, and the third backslash
   quotes the dollar sign.

   Every command language -- all UNIX shells, VMS DCL, DOS Batch, AOS/VS
   CLI, etc etc -- has similar rules. UNIX shell rules are probably the
   most complicated, since many printable characters -- not just one --
   are special there: dollar sign, single quote, double quote, backslash,
   asterisk, accent grave, number sign, ampersand, question mark,
   parentheses, brackets, braces, etc -- practically every
   non-alphanumeric character needs some form of quoting if it is to be
   taken literally. And to add to the confusion, the UNIX shell offers
   many forms of quoting, and many alternative UNIX shells are available,
   each using slightly different syntax.

1.11.2. Kermit's Quoting Rules

   Kermit's basic quoting rules are simple by comparison (there are, of
   course, additional syntax requirements for macro definitions, command
   blocks, function calls, etc, but they are not relevant here).

   The following characters are special in Kermit commands:

   Backslash (\)
          Introduces a variable, or the numeric representation of a
          special character, or a function, or other item for
          substitution. If the backslash is followed by a digit or by any
          of the following characters:

  x, o, d, m, s, f, v, $, %, &, :, {

          this indicates a special substitution item; otherwise the
          following character is to be taken literally (exceptions: \ at
          end of line is taken literally; \n, \b, and \n are special items
          in the OUTPUT command only).

   Semicolon (;)
          (Only when at the beginning of a line or preceded by at least
          one space or tab) Introduces a comment.

   Number sign (#)
          (Only when at the beginning of a line or preceded by at least
          one space or tab) Just like semicolon; introduces a comment.

   Question mark (?)
          (Only at the command prompt - not in command files or macros)
          Requests context-sensitive help.

   To force Kermit to take any of these characters literally, simply
   precede it by a backslash (\).

   Sounds easy! And it is, except when backslash also has a special
   meaning to the underlying operating system, as it does in DOS, Windows,
   and OS/2, where it serves as the directory separator in filenames such
   as:

  D:\K95\KEYMAPS\READ.ME

   Using our rule, we would need to refer to this file in Kermit commands
   as follows:

  D:\\K95\\KEYMAPS\\READ.ME

   But this would not be obvious to new users of Kermit software on DOS,
   Windows, or OS/2, and it would be annoying to seasoned ones. Thus
   MS-DOS Kermit and Kermit 95 go to rather extreme lengths to allow the
   more natural notation, as in:

  send d:\k95\keymaps\read.me

   The reason this is tricky is that we also need to allow for variables
   and other expressions introduced by backslash in the same command. For
   example, suppose \%a is a variable whose value is "oofa" (without the
   quotes). What does the following command do?

  send d:\%a

   Does it send the file named "oofa" in the current directory of the D:
   disk, or does it send a file named "%a" in the root directory of the D:
   disk? This is the kind of trouble we get into when we attempt to bend
   the rules in the interest of user friendliness. (The answer is: if the
   variable \%a has definition that is the name of an existing file, that
   file is sent; if a file d:\%a exists, it is sent; otherwise if both
   conditions are true, the variable takes precedence, and the literal
   filename can be forced by quoting: \\%a.)

   In Kermit 95 (but not MS-DOS Kermit), we also bend the rules another
   way by allowing you to use forward slash (/) rather than backslash (\)
   as the directory separator:

  send d:/k95/keymaps/read.me

   This looks more natural to UNIX users, and in fact is perfectly
   acceptable to the Windows 95/98/NT and OS/2 operating systems on the
   API level. BUT (there is always a "but") the Microsoft shell,
   COMMAND.COM, for Windows 95/98 and NT does not allow this notation, and
   therefore it can not be used in any Kermit command -- such as RUN --
   that invokes the Windows command shell AND your command shell is
   COMMAND.COM or any other shell that does not allow forward slash as
   directory separator (some alternative shells do allow this).

     NOTE: There exists a wide variety of alternative shells from third
     parties that do not have this restriction. If you are using a shell
     that accepts forward slash as a directory separator, you can stop
     reading right now -- UNLESS (there is always an "unless") you want
     your scripts to be portable to systems that have other shells. Also
     note that some Windows shells might actually REQUIRE forward slashes
     (instead of backslashes) as directory separators; we do not treat
     this situation below, but the treatment is obvious -- use slash
     rather backslash as the directory separator.

1.11.3. Passing DOS Filenames from Kermit to Shell Commands

   The following Kermit commands invoke the system command shell:

  RUN (and its synonyms ! and @)
  REDIRECT
  PIPE

   Each of these commands takes a shell command as an operand. These shell
   commands are not, and can not be, parsed by Kermit since Kermit does
   not know the syntax of shell commands, and so can't tell the difference
   between a keyword, a filename, a variable, a switch, or other item.
   Therefore the rules can not be bent since Kermit doesn't know where or
   how to bend them. To illustrate (using the regular Windows shell):

  run c:\\windows\\command\\chkdsk.exe

   works OK, but:

  run c:/windows/command/chkdsk.exe

   is not accepted by COMMAND.COM. But:

  run c:\windows\command\chkdsk.exe

   results in Kermit applying its quoting rules before sending the text to
   the shell. Since "w" and "c" are not in the list of backslash-item
   codes, the backslash means "take the following character literally".
   Thus, by the time this filename gets to the Windows shell, it has
   become:

  c:windowscommandchkdsk.exe

   which is probably not what you wanted. (If "w" and "c" were in the
   list, the results could be even stranger.) Even more confusing is the
   case where a directory or filename starts with one or more digits:

  run c:\123\lotus.exe

   in which "\123" is the Kermit notation for ASCII character 123, which
   happens to be left brace ({), resulting in "c:{lotus.exe".

   So when passing filenames to a Windows shell, always use double
   backslashes as directory separators, to ensure that the shell gets
   single backslashes:

  run c:\\windows\\command\\chkdsk.exe
  run c:\\123\\lotus.exe

   Similar problems might occur with the built-in EDIT, BROWSE, and FTP
   commands. These commands result in Kermit building a shell command
   internally to invoke the associated helper program; the form of this
   command might conflict with the form demanded by certain alternative
   shells.

1.11.4. Using Variables to Hold DOS Filenames

   Now to the next level. Suppose you want to write a script in which
   filenames are parameters, and therefore are stored in variables.
   Example:

  define \%f c:\windows\command\chkdsk.exe
  ...
  run \%f

   Obviously this won't work for the reasons just noted; the RUN command
   requires directory separators be coded as double backslashes:

  define \%f c:\\windows\\command\\chkdsk.exe
  ...
  run \%f

   This will work; no surprises here. However, if you had used ASSIGN
   rather than DEFINE, you might have been surprised after all; review
   pages 348-349 ofUsing C-Kermit (2nd Ed) for the difference
   between DEFINE and ASSIGN.

   We have said that any Kermit 95 or MS-DOS Kermit command that parses
   filenames itself -- SEND, for example -- does not require double
   backslashes since it knows it is parsing a filename. So since the
   following works:

  send c:\windows\command\chkdsk.exe

   Should the following also work?

  define \%f c:\windows\command\chkdsk.exe
  ...
  send \%f

   Answer: No. Why? Because \%f is evaluated "recursively", to allow for
   the possibility that its definition contains further variable
   references. This is true of all "backslash-percent-letter" (or -digit)
   variables, and also for array references. So \%f becomes
   c:\windows\command\chkdsk.exe, which becomes
   c:windowscommandchkdsk.exe.

   The trick here is to use the "other" kind of variable, that is
   evaluated only "one level deep" rather than recursively:

  define filename c:\windows\command\chkdsk.exe
  ...
  send \m(filename)

   Similarly if you want to prompt the user for a filename:

  ask filename { Please type a filename: }
   Please type a filename: c:\windows\command\chkdsk.exe
  send \m(filename)

1.11.5. Passing DOS Filenames as Parameters to Macros

   Suppose you want to pass a DOS filename containing backslashes as a
   parameter to a Kermit macro. This raises two issues:

    1. Parameters to macros are "just text" and so are fully evaluated
       before they are passed to the macro.
    2. Once inside the macro, the formal parameters \%1, \%2, ... \%9 are
       the type of variable that is evaluated recursively.

   Thus a DOS filename is ruined once in the act of parsing the macro
   invocation, and again when referring to it from within the macro. To
   illustrate, suppose "test" is a macro. Then in the invocation:

  test c:\mydir\blah.txt

   "c:mydirblah.txt" is assigned to \%1. However, if we double the
   backslashes:

  test c:\\mydir\\blah.txt

   "c:\mydir\blah.txt" is assigned to \%1. But then when you refer to \%1
   in the macro, it is evaluated recursively, resulting in
   "c:mydirblah.txt". To illustrate:

  define test echo \%1
  test c:\mydir\blah.txt
  c:mydirblah.txt
  test c:\\mydir\\blah.txt
  c:mydirblah.txt
  test c:\\\\mydir\\\\blah.txt
  c:\mydir\blah.txt

   Let's address each part of the problem separately. First, inside the
   macro. You can use the \fcontents() function to force a
   backslash-percent variable (such as a macro argument) to be evaluated
   one level deep instead of recursively, for example:

  define test echo { The filename is "\fcontents(\%1)"}

  test c:\mydir\blah.txt               ; We don't expect this to work
   The filename is "c:mydirblah.txt"   ; and it doesn't.
  test c:\\mydir\\blah.txt             ; But this does...
   The filename is "c:\mydir\blah.txt"

   Thus if the filename arrives inside the macro with single backslashes,
   the backslashes are preserved if you always refer to the parameter
   through the \fcontents() function.

   Now how to ensure that backslashes are not stripped or misinterpreted
   when passing a filename to a macro? This brings us back to what we
   learned in earlier sections:

    1. If it is a literal filename, either double the backslashes, or (if
       the filename is to be used only within Kermit itself and not passed
       to a DOS shell, or it is to be passed to an alternative shell that
       accepts forward slash as a directory separator), use forward slash
       instead of backslash as the directory separator.
    2. If it is a variable that contains a filename, make sure you use a
       macro-style variable name, rather than a
       backslash-percent-character name.

   Examples:

  define test echo \fcontents(\%1)
  define filename c:\mydir\blah.txt

  test c:\\mydir\\blah.txt  ; Literal filename with double backslashes
  c:\mydir\blah.txt

  test c:/mydir/blah.txt    ; Literal filename with forward slashes
  c:/mydir/blah.txt

  test \m(filename)         ; Variable
  c:\mydir\blah.txt

   But what if you don't like these rules and you still want to pass a
   literal filename containing single backslashes to a macro? This is
   possible too, but a bit tricky: turn command quoting off before
   invoking the macro, and then turn it back on inside the macro. Example:

  define test set command quoting on, echo \fcontents(\%1)

  set command quoting off
  test c:\mydir\blah.txt
  c:\mydir\blah.txt

   Upon return from the macro, command quoting is back on (since the macro
   turned it on).

   Obviously this trick can not be used if the filename is stored in a
   variable, since it prevents the variable from being evaluated.

1.11.6. Passing DOS File Names from Macro Parameters to the DOS Shell

   Now suppose you need to pass a DOS filename to a macro, and the macro
   needs to pass it, in turn, to the Windows shell via (say) Kermit's RUN
   command. This works too:

  define xrun run \fcontents(\%1)
  xrun c:\\windows\\command\\chkdsk.exe

   (or you can use the SET COMMAND QUOTING OFF / ON technique described
   above to avoid the double backslashes.) But..

  xrun c:/windows/command/chkdsk.exe

   does not work if the Windows shell does not recognize "/" as a
   directory separator. If there is a chance that a filename might be
   passed to the macro in this form, the macro will need to convert it to
   a form acceptable to the shell:

  define xrun run \freplace(\fcontents(\%1),/,\\)

   Here we replace all occurrences (if any) of "/" in the argument with
   "\" prior to issuing the RUN command. Of course, in order to specify
   "\" as a literal character in the \freplace() argument list, we have to
   double it.

1.11.7. Passing DOS Filenames to Kermit from the Shell

   As noted in the manual, the \&@[] array contains Kermit's command-line
   arguments. Suppose one of these arguments, say \&@[3], is a DOS
   filename such as C:\FOO\BAR\BAZ\OOFA.TXT. (Note: In C-Kermit 7.0 and
   K95 1.1.18 and later, command-line arguments after "=" or "--" are also
   available in the top-level \%1..9 variables; see Section 7.5.)

   Of course you can eliminate any problems by using forward slashes
   rather than backslashes in the filename, but sometimes this is not
   possible, as when the Kermit command line is being generated by another
   program than can only generate "native" format DOS filenames.

   As noted in the manual, "\%x" variables and \&x[] arrays are always
   evaluated "all the way" (recursively). If the contents of one of these
   variables contains backslashes, this causes another level of
   evaluation.

   There is another kind of variable, which is evaluated only "one level
   deep". You can use this to prevent interpretation of the backslashes in
   the filenames. Example:

  assign filename \fcontents(\&@[3])  ; Transfer contents
  ...
  send \m(filename)

   Or, more simply:

  send \fcontents(\&@[3])

1.12. Debugging

   The debug log is produced when you give a "log debug" command. This is
   normally done at the request of the Kermit help desk, for forwarding to
   the Kermit developers for analysis as a last resort in troubleshooting
   problems. (Last resort because it can grow quite huge in a very short
   time.) In cases where timing information is critical to understanding a
   problem, you can tell C-Kermit to put a timestamp on each debug log
   line by giving the command:

  SET DEBUG TIMESTAMP ON

   At any time before or after activating the debug log (SET DEBUG
   TIMESTAMP OFF turns off timestamping). Timestamps can be turned off and
   on as desired while logging. Obviously, they increase the size and
   growth rate of the log significantly, and so should be used sparingly.
   Timestamps are of the form hh:mm:ss.xxx, where .xxx is thousands of a
   second (but is included only on platforms that include this feature).

1.13. Logs

   In UNIX C-Kermit and in K-95, you can now direct any log to a pipe.
   This not only lets you send your logs to places other than disk files,
   but also lets you customize them to any desired degree.

   LOG { DEBUG, PACKETS, SESSION, TRANSACTION, CONNECTION } { file, pipe }
          ...
          A "pipe" is the name of a command, preceded by a vertical bar.
          If the pipe contains any spaces, it must be enclosed in braces.

   Here are some examples for UNIX (always remember the importance of
   getting the UNIX shell quoting rules right):

   LOG TRANSACTIONS |lpr
          This sends the transaction log to the default UNIX printer,
          rather than to a file (use "lp" rather than "lpr" if necessary).

   LOG TRANSACTIONS {| myfilter > t.log}
          For those who don't like the format of the transaction log, or
          want to extract certain information from it; write your own
          output filter.

   LOG SESSION {| lpr -Plaserwriter}
          This sends the session log to a specific UNIX printer, rather
          than to a file. Note the braces around the pipeline. These are
          required because it contains spaces.

   LOG DEBUG {| tail -100 > debug.log}
          This causes the debug log file to contain only the final 100
          lines. Suppose C-Kermit crashes under some unpredictable
          circumstances, and you need a debug log to catch it in the act.
          But the debug log can grow to huge proportions very quickly,
          possibly filling up the disk. Piping the debug log through
          "tail" results in keeping only the last 100 lines (or other
          number of your choice).

   LOG DEBUG {| grep "^TELNET" > debug.log}
          This one shows how to log only Telnet negotiations. Piping the
          debug log through grep or egrep lets you log only specific
          information, rather than everything. "man grep" for further
          info.

   LOG DEBUG {| gzip -c > debug.log.gz}
          Creates a full debug log, but compressed by gzip to save space.

   LOG PACKETS {| tr "\\01" "X" | cut -c9- > packet.log}
          This one writes the regular packet log, but translates the
          Ctrl-A that starts each packet to the letter "X" and removes the
          s-nn-nn- notation from the beginning of each line. Note the
          double backslash (normal Kermit quoting rules). "man tr" and
          "man cut" for further info.

   See Section 2.12 for information about the new connection log.

1.14. Automatic File-Transfer Packet Recognition at the Command Prompt

   Beginning in version 7.0, C-Kermit can recognize Kermit (and in some
   cases also Zmodem) file-transfer packets while at its command prompt.
   This is convenient (for example), if you escaped back from a remote
   Kermit program and told the local Kermit program to send a file, but
   forgot to tell the remote Kermit program to receive it (and the local
   Kermit did not have the "send a Kermit receive command" feature
   available). This feature is controlled by the following command:

   SET COMMAND AUTODOWNLOAD { ON, OFF }
          When ON, which is the default, the command parser recognizes
          Kermit packets when Kermit is in remote mode. An S packet makes
          it go into receive mode, an I packet makes it go into server
          mode. When OFF, packet recognition is disabled and the behavior
          when a packet is received at the command prompt is as it was in
          C-Kermit 6.1 and earlier (namely to print an error message).

   COMMAND AUTODOWNLOAD is the command-mode equivalent of TERMINAL
   AUTODOWNLOAD, which is effective during CONNECT mode.

1.15. The TYPE Command

   The TYPE command now accepts a selection of optional switches
   ( Section 1.5), and also sets several variables.

   Syntax: TYPE [ switches... ] filename

   Variables:

   \v(ty_ln)
          Line number of current line (during TYPE command; see /PREFIX)

   \v(ty_lc)
          Line count of file most recently TYPEd.

   \v(ty_mc)
          Match count of file most recently TYPEd (see /MATCH).

   Switches:

   /PAGE
          If /PAGE is included, Kermit pauses at the end of each screenful
          and issues a "more?" prompt. You may press the space bar to view
          the next page (screenful), or press "q" or "n" to return to the
          C-Kermit prompt. If this switch is given, it overrides the
          COMMAND MORE-PROMPTING setting for this command only. If it is
          not given, paging is according to COMMAND MORE-PROMPTING.

   /NOPAGE
          Do not pause at the end of each screenful; show the whole file
          (or all selected lines) at once. If this switch is given, it
          overrides the COMMAND MORE-PROMPTING setting for this command
          only. If it is not given, paging is according to COMMAND
          MORE-PROMPTING.

   /HEAD[:n]
          Only show the first n lines of the file (where n is a number).
          If n is omitted, 10 is used.

   /TAIL[:n]
          Only show the last n lines of the file (where n is a number). If
          nis omitted, 10 is used. Note: /HEAD and /TAIL can't be
          combined; if you give both switches, only the most recent one is
          used.

   /MATCH:pattern
          Only type lines from the file that match the given pattern (see
         Section 4.9.1 for pattern notation). UNIX users familiar
          with grep should note a significant difference: there is no
          implied "*" at the beginning and end of the pattern. Thus:

  TYPE /MATCH:foo    Lists lines whose entire contents are "foo".
  TYPE /MATCH:foo*   Lists lines that start with "foo".
  TYPE /MATCH:*foo   Lists lines that end with "foo".
  TYPE /MATCH:*foo*  Lists lines that have "foo" anywhere in them.

          /HEAD and /TAIL apply after /MATCH, so "type /tail:20 /match:x*"
          shows the last 20 lines in the file that start with "x".

   /PREFIX:string
          Print the given string at the beginning of each line. The string
          may be a constant, a variable, or a quoted variable. If it's an
          unquoted variable, its value at the time the TYPE command was
          given is used as a constant. If it is a quoted variable, it is
          re-evaluated for each line; a useful variable for this context
          is \v(ty_ln) (the line number of the current line being typed).
          If the prefix is to include spaces, it must be enclosed in
          braces. Examples:

        type /prefix:{oofa.txt: } /match:*thing* oofa.txt
                Prints all lines in oofa.txt that contain "thing" with the
                filename itself as the prefix (similar to UNIX grep).

        type /prefix:{\v(time). } oofa.txt
                Prefixes each line of oofa.txt with the time at which the
                TYPE command was given (one backslash)

        type /prefix:{\\v(time). } oofa.txt
                Prefixes each line of oofa.txt with the time at which that
                line is being typed (two backslashes).

        type /prefix:{\\v(ty_ln). } oofa.txt
                Prefixes each line of oofa.txt with its line number.

        type /prefix:{\\flpad(\\v(ty_ln),4). } oofa.txt
                Same as the previous example, except the line number is
                right-adjusted in a 4-column field.

   /WIDTH[:n]
          Truncates each line at column n (which must be a number) prior
          to printing it. This option can be used for long lines when you
          don't want them to wrap. If nis omitted, your current screen
          width is used.

   /COUNT
          Counts lines and -- if /MATCH was included, matches -- but does
          not print any lines from the file. The line and match count is
          shown at the end, and the variables \v(ty_lc) and \v(ty_lm) are
          set accordingly.

   SET OPTIONS TYPE { /PAGE, /NOPAGE, /WIDTH:n }
          Sets the paging default for TYPE commands, which can be
          overridden in any particular TYPE command by including the
          desired switch.

   If a TYPE command is given with no switch, and no SET OPTIONS TYPE
   selection is in effect, paging is according to your COMMAND
   MORE-PROMPTING setting (SHOW COMMAND).

1.16. The RESET Command

   The RESET command, added in 7.0, closes all open files and logs, but
   does not affect the open connection (if any).

1.17. The COPY and RENAME Commands

   As of C-Kermit 7.0, in the UNIX version only, the COPY and RENAME
   commands are built in and do not call the underlying platform's COPY or
   RENAME command. This allows them to work in "NOPUSH" versions and other
   circumstances where it can't access system commands, and it allows file
   copying and renaming to be done portably in scripts. The
   characteristics of the built-in COPY or RENAME include:
     * It fails if the source file is a directory or is wild or lacks read
       access.
     * It fails if the source file is the destination file.
     * It allows the destination file to be a directory, in which case the
       source file is copied (or renamed) into it with the same name.
     * It overwrites an existing destination file if its permission
       allows.
     * It sets the new file's permission according to umask but also
       carries forward the source file's execute permission bits if the
       destination file did not already exist.
     * It fails if interrupted by Ctrl-C.
     * Upon error, it prints an appropriate message.
     * It returns standardized error codes that can be tested by IF
       SUCCESS / FAIL.

   These commands now also accept the following switches:

  /LIST (/LOG, /VERBOSE)    = Print "file1 => file2 (OK)" (or error message).
  /NOLIST (/NOLOG, /QUIET)  = Don't print anything (except error messages).

   /NOLIST is the default.

   The same built-in code is used by the UNIX C-Kermit server to execute
   REMOTE COPY commands (except in this case no switches are available).

   The COPY command also accepts the following additional switches. When
   any of these are given (and they can be used in any combination except
   /SWAP and /APPEND), some of the checks listed above are relaxed, and
   thus it might be possible to get into trouble in certain cases, e.g.
   when the source and target files are the same file:

  /APPEND                   = Append source file to destination file.
  /SWAP-BYTES               = Swap bytes (see Section 6.6.5).
  /FROMB64                  = Decode the source file from Base64 encoding.
  /TOB64                    = Encode the target file in Base64.

   Base64 is the encoding commonly used for enclosures in Internet email.

1.18. The MANUAL Command

   The MANUAL command can be used to access the appropriate Kermit manual
   or other manual. The general syntax is:

   MANUAL [ string ]
          If the string is omitted, C-Kermit asks the underlying system to
          access the C-Kermit manual using whatever method is appropriate
          for the system.

   The specific action depends on the system. In UNIX, a "man" command is
   issued; "kermit" is the default argument but other manual topics may be
   specified. If the "man" command allows index or string searching, the
   appropriate syntax may be included.

   In Kermit 95, the MANUAL command brings up the HTML online K95 manual.

   In VMS and elsewhere, "man" is simply translated to "help", with a
   default argument of "kermit"; other and/or additional arguments may be
   included according to the definition of the system's "help" command.

   Correct operation of the "man" command in C-Kermit depends on the
   appropriate man page or help topic having been installed in the right
   place with the right permissions and format.

1.19. String and Filename Matching Patterns

   A pattern is a string that includes special notation for matching
   classes or sequences of characters. C-Kermit 7.0 / K95 1.1.19 supports
   patterns in several places:

     * Filenames ( Section 4.9)
     * SWITCH case labels ( Section 7.18)
     * The new IF MATCH statement ( Section 7.4)
     * TYPE /MATCH ( Section 1.15)
     * SET FILE TEXT-PATTERNS and BINARY-PATTERNS ( Section 4.3)
     * The \fsearch() and \farraylook() functions (Sections 7.3 and
      7.10.7)
     * The \fpattern() function used with [M,RE]INPUT ( Section 7.1)

   Patterns are also called wildcards, especially when used for filename
   matching. C-Kermit's pattern syntax is explained in Section 4.9.1,
   and also by the HELP WILDCARDS command.

1.20. Multiple Commands on One Line

   As of C-Kermit 7.0, commands can be grouped together on one line by
   separating the commands with commas and enclosing the list in braces.
   For example:

  C-Kermit> { echo One, echo Two, echo Three }
  C-Kermit> do { echo One, echo Two, echo Three }

   Command lists can be nested:

  [ do ] { echo One, echo Two, if true { echo A, echo B}, echo Three }

   and the END command works as it does in macros:

  [ do ] { echo One, echo Two, if true end, echo Three }

   The "one line" stricture is, of course, pliant to line-continuation
   conventions, namely that lines ending in hyphen (-) or left brace ({)
   are to be continued. Thus the first example can also be rendered:

  [ do ] {
      echo One
      echo Two
      echo Three
  }

   (the "do" is optional).

1.21. What Do I Have?

   C-Kermit can be built for hundreds of different platforms with
   practically countless configuration options. Certain commands might not
   be available in certain configurations, etc. Even on the same platform,
   different builds are possible: "maximum functionality", "minimum size",
   "maximum performance", and so on. You can find out a lot about the
   configuration of your C-Kermit program with the SHOW FEATURES command.
   Of course, a lot of what it says, especially in the bottom part, might
   seem like gibberish, but can be deciphered with a Rosetta Stone (such
   as the C-Kermit source or theckccfg.txt file). In any case, the
   output from SHOW FEATURES might easily explain why some expected
   feature is missing, or some buffer is smaller than expected. Here's a
   sample of the bottom section for the SunOS version:

C-Kermit 7.0.196, 1 Jan 2000

Major optional features included:
 Network support (type SHOW NET for further info)
 Telnet Kermit Option
 Hardware flow control
 External XYZMODEM protocol support
 Latin-1 (West European) character-set translation
 Latin-2 (East European) character-set translation
 Cyrillic (Russian, Ukrainian, etc) character-set translation
 Greek character-set translation
 Hebrew character-set translation
 Japanese character-set translation
 Unicode character-set translation
 Pseudoterminal control
 REDIRECT command
 RESEND command
 Fullscreen file transfer display
 Control-character unprefixing
 Streaming
 Autodownload

Major optional features not included:
 No Kerberos(TM) authentication
 No SRP(TM) (Secure Remote Password) protocol
 No Secure Sockets Layer (SSL) protocol
 No Transport Layer Security (TLS) protocol
 No encryption
 No X Windows forwarding

Host info:
 Machine:    sun4m
 Model:      (unknown)
 OS:         SunOS
 OS Release: 4.1.3_U1
 OS Version: 4

Target: sunos41gsc
GCC version: 2.7.2
Compiled Dec 31 1999 10:38:54, options:
 __GNUC__ __STDC__ _POSIX_JOB_CONTROL _SC_JOB_CONTROL ARRAYREFLEN=1024 BIGBUFOK
 BROWSER BSD4 CK_ANSIC CK_APC CK_AUTODL CK_CURSES CK_DNS_SRV CK_ENVIRONMENT
 CK_FAST CK_LOGIN CK_MKDIR CK_NAWS CK_PCT_BAR CK_PERMS CK_RECALL CK_RTSCTS
 CK_SPEED CK_TIMERS CK_TMPDIR CK_TTGWSIZ CK_TTYFD CK_WREFRESH CKEXEC
 CKFLOAT=double CKGHNLHOST ckmaxfiles=64 CKMAXOPEN=64 CKMAXPATH=1023 CKREALPATH
 CKREGEX CKSYSLOG CKTUNING CMDBL=32763 CMDDEP=64 CONGSPD DCMDBUF DIRENT DYNAMIC
 FNFLOAT FORDEPTH=32 GFTIMER HADDRLIST HDBUUCP IFDEBUG IKS_OPTION IKSDB
 IKSDCONF INBUFSIZE=32768 INPBUFSIZ=4096 MAC_MAX=16384 MACLEVEL=128 MAXDDIR=32
 MAXDNUMS=4095 MAXGETPATH=128 MAXTAKE=54 MAXWLD=102400 MSENDMAX=1024 NETCMD
 NETCONN NETPTY NOKVERBS NOSETBUF OBUFSIZE=32768 PARSENSE PATTERNS PIPESEND
 RENAME RLOGCODE SAVEDUID SELECT SIG_V SOL_SOCKET sparc STREAMING sun SUNOS4
 SYSTIMEH TCPSOCKET TIMEH TLOG TNCODE TTLEBUF TTSPDLIST UIDBUFLEN=256 UNIX
 UNPREFIXZERO USE_LSTAT USE_MEMCPY VNAML=4096 WHATAMI XFRCAN Z_MAXCHAN=46
 z_maxchan=46 ZXREWIND

 byte order: big endian

 sizeofs: int=4 long=4 short=2 char=1 char*=4 float=4 double=8

 floating-point: precision=16 rounding=1

   Without going into detail about what all the notation means, notice a
   couple things:

     * The Options section shows symbols ("macros") in effect during
       compilation, together with their values (for those that have
       values). The options are listed in alphabetical order to make any
       particular option easier to find.
     * MAXWLD is the maximum number of files that a wildcard can expand
       to.
     * Anything starting with "NO" is a feature (or something other than a
       feature) that has been deliberately "compiled out", or omitted.
     * Important items for script writers include: CMDBL=32763 (the size
       of the command buffer and therefore the maximum length for a macro
       or variable definition; CMDDEP=64 (the limit on recursion depth);
       FORDEPTH=32 (the nesting limit on FOR loops); INBUFSIZE=32768 (the
       size of the INPUT command circular buffer); MAC_MAX=16384 (the
       maximum number of macros), etc.

   See theckccfg.txt file for details.

1.22. Generalized File Input and Output

   C-Kermit 7.0 adds a new generalized I/O system for stream files,
   augmenting (and to some extent, overlapping with) the older OPEN, READ,
   WRITE, and CLOSE commands. In the new file i/o system, which can be
   used simultaneously with the old one, all commands are grouped together
   under the new FILE keyword, and some related functions and variables
   are added.

1.22.1. Why Another I/O System?

   The well-known LOG, OPEN, READ, WRITE, and CLOSE commands have the
   following restrictions:

    1. Only one READ file and one WRITE file can be open at a time.
    2. The READ and WRITE commands are strictly line oriented.
    3. These commands can not be used with binary files.
    4. They do not support read/write access or random access.
    5. The syntax is a bit counterintuitive for programmers.

   The new file i/o system allows multiple files to be open at once, in
   any desired combination of modes (read/write/append) supported by the
   operating system, for line, block (record), or character i/o, for
   sequential or random access, using consistent syntax and conventions.

   The new system, however, does not replace the old one, since the old
   system still must be used for:

    1. The session, packet, debug, transaction, and connection logs.
    2. Reading and writing commands rather than files.
    3. Existing scripts.

   The new system works only with regular files, not with commands or
   pipes or mailboxes or pseudoterminals. No special provisions are made
   in the FILE commands for handling devices or network connections, nor
   for preventing you from trying to open them; if the underlying
   operating system treats them like regular stream disk files, the FILE
   commands (except, of course SEEK, REWIND, and COUNT) might work with
   them. (In C programming terms, the FILE commands are, at present,
   nothing more than a front end to fopen() / fread() / fwrite() /
   fclose() and friends, which are a portable API to sequential files, but
   this might change in the future for platforms like VMS and VOS that
   have more complicated file systems.)

   Definitions:

   Channel
          A number assigned to a file when it is opened, by which it must
          be referred to in all input/output operations.

   Read/Write Pointer
          The current position in an open file, expressed as the 0-based
          byte count from the beginning.

1.22.2. The FILE Command

   FILE keyword [ switches ] channel [ data ]
          The keyword specifies the function: FILE OPEN, FILE READ, FILE
          WRITE, FILE CLOSE, etc. For convenience (and for familiarity to
          C programmers), the two-word FILE commands can be shortened to
          the single words FOPEN, FREAD, FWRITE, FCLOSE, and so on.
          Switches are optional, and modify or amplify the requested file
          function.

   As in C, Fortran, and other programming languages, open files are
   referred to by "channels", integers such as 0, 1, 2, 3, and so on. A
   channel number is assigned when you open a file. The number of
   available channels depends on the underlying operating system, and can
   be seen in the variable:

  \v(f_max)

   or by giving the FILE LIST (FLIST) command. Channels are discussed in
   greater detail in Section 1.22.4.

   FILE command errors can be caught with IF FAIL after the FILE command.
   In addition, the \v(f_error) variable is set to the completion code of
   the command: 0 if no error, or a negative number if there was an error.
   The error codes are listed in Section 1.22.5.

   The command to open a file is:

   FILE OPEN [ switches ] variable filename
          Opens a file for the type of access specified by the switches,
          or for read-only access if no switches are given. Upon success,
          a channel number is assigned to this file and stored in the
          given variable so you can refer to the open file in subsequent
          i/o commands. If the file can not be opened, the FILE OPEN
          command fails. Synonym: FOPEN.

   The FILE OPEN switches are:

   /READ
          Open the file for read access. If no switches are given, /READ
          is assumed. If the file does not exist or can't be opened for
          read access, the FILE OPEN command fails.

   /WRITE
          Allow writing. If a file of the same name already exists, it is
          overwritten unless /READ or /APPEND is also included. If a file
          of the given name does not exist, it is created.

   /APPEND
          Equivalent to /WRITE, except that if the file exists, it is not
          destroyed. The read/write pointer is set to the end of the file,
          so unless you change it with FILE SEEK or REWIND (see below),
          the first FILE WRITE command adds to the end of the file,
          preserving what was there already. If /WRITE is also given, it
          is ignored.

   /BINARY
          Open the file in "binary" mode, rather than text mode. This
          switch is meaningless (but still can be used) in UNIX. In VMS,
          Windows, and OS/2, it inhibits end-of-line processing and
          conversion, and so should be used for binary files and/or files
          that are to be accessed in record or character mode rather than
          line by line.

   The variable for the channel number can be any kind of variable: the
   \%x kind, a macro name, or an array element. But it must be a variable,
   not a number -- C-Kermit assigns the channel number; you can't tell it
   what number to use.

   Example:

  FILE OPEN \%c oofa.txt                  ; Open oofa.txt for reading.
  IF FAIL exit 1 Can't open oofa.txt      ; Always check to see if it worked.
  ECHO oofa.txt: channel = \%c

   If the file oofa.txt is opened successfully, a channel number is
   assigned to the variable \%c. Here's another example using a macro name
   for the channel number:

  FILE OPEN channel oofa.txt              ; Open oofa.txt for reading.
  IF SUCCESS ECHO oofa.txt: channel = \m(channel)

   Switches can be combined when it makes sense and the underlying
   operating system allows it. For example, to open a file in binary mode
   for reading and writing (sometimes called "update"):

  FILE OPEN /READ /WRITE /BINARY \%c budget.db

   Some combinations might be allowed, others not. For example /READ
   /APPEND will usually not be allowed. /WRITE /APPEND is treated as
   /APPEND.

   A major advantage of the new system over the older one is that you can
   have multiple files open at once. Suppose, for example, that you want
   to open all the files in a certain directory at once:

  .\%n := \ffiles(/usr/olga*,&f)          ; Get file list into array.
  if ( > \%n \v(f_max) ) {                ; Make sure there aren't too many.
      exit 1 {\v(dir): \%n = Too many files}
  }
  declare \&c[\%n]                        ; Make array for channel numbers.
  for \%i 1 \%n 1 {                       ; Loop to open every file...
      file open \&c[\%i] \&f[\%i]         ; Try to open this one
      if fail exit 1 Open error: \&f[\%i] ; Check for failure
  }

   If this loop completes successfully, the \&c[] array will contain \%n
   channel numbers of open files in elements 1 through \%n.

   Any file that you open with FILE OPEN stays open until Kermit exits, or
   you close it explicitly. The command to close a file is:

   FILE CLOSE { ALL, channel }
          If a channel number is given and the channel refers to an open
          file, the file is closed and the channel is freed for reuse; if
          the channel does not refer to an open file, an error message is
          printed and the command fails. If ALL is specified instead of a
          specific channel, all files opened with FILE OPEN are closed and
          if all open files were closed successfully (even if no files
          were open), the command succeeds; if any open file could not be
          closed, the command fails; however, all open files that could be
          closed are still closed. Synonym: FCLOSE.

   FILE CLOSE might fail because, for example, the disk filled up or a
   quota was exceeded. Example:

  fopen /write \%c new.txt                ; Open new.txt for writing.
  if fail exit 1                          ; Check for error.
  fclose \%c                              ; Close the file we just opened.

   This creates a 0-length file called new.txt.

   Note that FILE OPEN /WRITE (without /READ or /APPEND) always creates a
   new file, and therefore destroys any file with the same name that might
   already exist (assuming you have permission to delete it). To avoid
   overwriting existing files, simply check first:

  if exist new.txt exit 1 {Fatal - new.txt already exists}
  fopen /write \%c new.txt
  if fail ...

   The next two commands give information about open files:

   FILE STATUS channel
          Tells the name of the file, if any, open on the given channel
          and the switches it was opened with. The read/write pointer is
          also shown; this is where the next read or write will occur;
          "[EOF]" is shown if the current position in the open file is the
          end -- i.e. the next read will fail if the file was opened in
          /READ mode; the next write will add material to the end. The
          current line number (0-based) is also shown if known. The FILE
          STATUS command succeeds if the channel is open, and fails if
          there is no open file on the given channel, or if the channel
          number is invalid or out of range. Synonym: FSTATUS.

   FILE LIST
          Lists the channel number and name of each open file, along with
          its OPEN modes (R, W, A, B, RW, etc) and its current read/write
          pointer or "[EOF]" if it is at the end. Also tells the number of
          files currently opened with FILE OPEN, plus the maximum number
          of open files allowed by the system and the maximum number
          allowed for FILE OPEN. Synonym: FLIST.

   Next come the commands for reading and writing files:

   FILE READ [ switches ] channel [ variable ]
          Reads data from the file on the given channel number into the
          variable, if one was given; if no variable was given, the result
          is printed on the screen. IMPORTANT: The variable should
          normally be a macro name rather than a \%x or \&x[] variable if
          you want backslash characters in the file to be taken literally
          (see pp.408-412 ofUsing C-Kermit for an explanation; you
          can also read into a \%x or \&x[] variable, but then you must
          remember to protect future references to by \fcontents() if you
          don't want C-Kermit to process any backslashes it might
          contain). The desired amount of data (according to the switches)
          is read from the file at the current read/write pointer, and
          upon completion the read/write position is updated to first byte
          after the data that was read, no matter what switches were
          given. Synonym: FREAD.

   FILE WRITE [ switches ] channel text
          Writes the given text to the file on the given channel number.
          The text, of course, can be literal text or a variable, or any
          combination. If the text might contain leading or trailing
          spaces, it must be enclosed in braces if you want to preserve
          them. Synonym: FWRITE.

   Before proceeding, a caution about the NUL character. C-Kermit is so
   named because it is a Kermit program written in the C language. In C,
   character strings are represented as a sequence of non-NUL bytes
   terminated by a NUL byte (a byte in which all bits are 0). Thus a C
   string can not contain NUL bytes; it always ends with the first NUL
   byte. C-Kermit variables are implemented as C strings and therefore
   can't contain NUL bytes either, so the FILE READ and FILE WRITE
   commands do not handle files or strings that contain NUL bytes, except
   when the /CHARACTER switch is included with the FILE READ or WRITE
   command, or when /LPAD:0 or /RPAD:0 is given with the FILE WRITE
   command; these switches are explained below.

   Also note that Kermit can not be used read or write binary numbers in
   the machine's internal format (integer or floating-point); in general,
   numbers can be processed only when represented as numeric or
   floating-point strings.

   FILE READ switches are:

   /LINE
          Specifies that a line of text is to be read. A line is defined
          according to the underlying operating system's text-file format.
          For example, in UNIX a line is a sequence of characters up to
          and including a linefeed, or the end of the file, which ever
          comes first. The line terminator (if any) is removed before
          assigning the text to the variable. If no switches are included
          with the FILE READ command, /LINE is assumed. Normally this
          switch should not be used with files opened in /BINARY mode (but
          nothing prevents it either).

   /SIZE:number
          Specifies that the given number of bytes (characters) is to be
          read. The actual number of bytes returned will be less if the
          end of file is reached (or a NUL byte is encountered). For
          example, if a file is 514 bytes long, FILE READ /SIZE:512
          returns 512 bytes the first time and 2 bytes the second time.
          FILE READ /SIZE provides a kind of "record i/o" for files that
          do not necessarily contain lines. The resulting block of
          characters is assigned to the variable without any editing.
          Synonym: /BLOCK.

   /CHARACTER
          Equivalent to /SIZE:1. If FILE READ /CHAR succeeds but the
          variable is empty, this indicates a NUL byte was read. Synonym:
          BYTE.

   FILE WRITE switches are:

   /LINE
          Specifies that an appropriate line terminator is to be added to
          the end of the text. If no switches are included, /LINE is
          assumed.

   /SIZE:number
          Specifies that the given number of bytes (characters) is to be
          written. If the given text is longer than the requested size, it
          is truncated; if is shorter, it is padded according /LPAD and
          /RPAD switches. Synonym: /BLOCK.

   /LPAD[:value]
          If /SIZE was given, but the text is shorter than the requested
          size, the text is padded on the left with sufficient copies of
          the character whose ASCII value is given to write the given
          length. If no value is specified, 32 (the code for Space) is
          used. The value can also be 0 to write the indicated number of
          NUL bytes. If /SIZE was not given, this switch is ignored.

   /RPAD[:value]
          Like LPAD, but pads on the right.

   /CHARACTER
          Specifies that one character should be written. If the text is
          empty or not given, a NUL character is written; otherwise the
          first character of text is given. Synonym: /BYTE.

   /STRING
          Specifies that the text is to be written as-is, with no
          terminator added.

   Here's an example in which we copy a text file line by line:

  file open /read \%c oofa.txt            ; Open input file
  if fail exit 1 Can't open input file    ; Check that it's open
  file open /write \%d new.txt            ; Open output file
  if fail exit 1 Can't open output file   ; Check
  while true {                            ; Loop to copy lines
      file read /line \%c line            ; Read a line
      if fail break                       ; Assume failure = end of file
      file write /line \%d {\m(line)}     ; Write the line to output file
      if fail exit 1 Write failure        ; Failure here is fatal
  }
  file close \%c                          ; Close the two files
  file close \%d

   Note that since /LINE is the default for both FILE READ and FILE WRITE,
   it can be omitted as in the following example, where we also use the
   short names for the FILE commands.

  fopen /read \%c oofa.txt                ; Open input file
  if fail exit 1 Can't open input file    ; Check that it's open
  fopen /write \%d new.txt                ; Open output file
  if fail exit 1 Can't open output file   ; Check
  while true {                            ; Loop to copy lines
      fread \%c line                      ; Read a line
      if fail break                       ; Assume failure = end of file
      fwrite \%d {\m(line)}               ; Write the line to output file
      if fail exit 1 Write failure        ; Failure here is fatal
  }
  fclose \%c                              ; Close the two files
  fclose \%d

   Here's the same example using "record i/o" (the open and close
   sequences are are omitted since they are the same as above). The result
   is the same, but execution is much faster:

  while true {                            ; Loop to copy blocks
      fread /size:512 \%c block           ; Read a block into \%a
      if fail break                       ; Assume failure = end of file
      fwrite /string \%d {\m(block)}      ; Write the block to output file
      if fail exit 1 Write failure        ; Failure here is fatal
  }

   Although record i/o is faster, it should not be used in line-oriented
   applications, since it returns arbitrary chunks of the file to your
   script, rather than lines. In this example, FWRITE /STRING is used
   rather than FWRITE /SIZE:512 to avoid the last output block being
   padded beyond the original file's length.

   A file can also be copied character by character, but this is much
   slower than line i/o and VERY much slower than block i/o:

  while true {                            ; Loop to copy blocks
      fread /char \%c c                   ; Read a character into c
      if fail break                       ; Assume failure = end of file
      fwrite /char \%d {\m(c)}            ; Write character to output file
      if fail exit 1 Write failure        ; Failure is fatal
  }

   Although character i/o is slow, it is the only way to process files
   that contain NUL characters (i.e. bytes composed of only zero bits). In
   the example above, when "fread /char \%c c" returns a NUL, the c
   variable is empty. But since the FREAD /CHAR command did not fail, we
   know the result was really a NUL. FWRITE /CHAR, when given an empty
   variable (or no variable at all) writes a NUL. Thus the loop above will
   copy any file at all (very slowly). In non-copying applications, NULs
   are detected like this:

  fread /char \%c c
  if fail (do something)
  if not def c (a NUL byte was read)

   Finally some advanced file operations:

   FILE FLUSH channel
          For output files only: commits all previous writes to disk, in
          case the computer was buffering them. Synonym: FFLUSH.

   FILE COUNT [ { /BYTES, /LINES, /LIST, /NOLIST } ] channel
          By default, or if the /BYTES switch is given, counts the bytes
          in the file, if any, open on the given channel. If the /LINES
          switch is given, counts lines in the file. If the /LIST switch
          is given, the result is printed. If the /NOLIST switch is given,
          the result is not printed. /QUIET is a synonym for /NOLIST. If
          neither /LIST nor /NOLIST is given, the result is printed if the
          command is given at top level, i.e. not from a command file or
          macro. In all cases, the result of the most recent FILE COUNT
          command is stored in the variable \v(f_count). Note that FILE
          COUNT /LINE works (and can only work) by reading the entire
          file; expect it to take some time if the file is large. Synonym:
          FCOUNT.

   FILE REWIND channel
          Moves the read/write pointer to the beginning of the file.
          Equivalent to FILE SEEK channel 0. Synonym: FREWIND.

   FILE SEEK [ switches ] channel { [{+,-}]number, LAST, EOF }
          Moves the read/write pointer for the file on this channel to the
          given position, which may be a byte (character) number or a line
          number, expressed in either absolute or relative terms.
          Switches:

        /BYTE
                The number given is a byte number. Synonym: /CHARACTER.

        /LINE
                The number given is a line number.

        /ABSOLUTE
                The number given is absolute.

        /RELATIVE
                The number given is relative to the current position.

          By default, or if the /BYTE switch is given, the number is a
          byte number (0 = first byte). If /LINE is given, the number is a
          line number (0 = first line). EOF means to move to the end of
          the file. LAST means to move to the last line or character of
          the file, depending on whether it's a line or character seek.

          If neither the /RELATIVE nor the /ABSOLUTE switch is given, then
          if a signed number is given, the motion is relative to the
          current position. An expression that evaluates to a negative
          number is not considered signed for this purpose; that is, a
          sign (+ or -) must be included as the first character of the
          number in the command itself to force a relative seek (in the
          absence of /RELATIVE or /ABSOLUTE).

          If the number has no sign, or if the /ABSOLUTE switch is given,
          the number represents an absolute position (relative to the
          beginning of the file). Subsequent FILE READs or WRITEs will
          take place at the new position.

          If the read/write pointer is placed after the end of the file, a
          subsequent FILE READ will fail, but a FILE WRITE will succeed
          (possibly creating a file with "holes"). If a FILE SEEK /BYTE
          command is given, the current line becomes unknown (unless the
          position is 0) and subsequent FILE SEEK /RELATIVE /LINE commands
          will fail until the next non-relative FILE SEEK /LINE command is
          given. Synonym: FSEEK.

   An absolute FILE SEEK to a negative position fails silently, as does a
   relative seek to a position before the beginning of the file.

   A caution about relative SEEKs: remember that the number is relative to
   the current position. Whenever you read or write, this changes the
   position. In each of the following examples, assume the file open on
   channel \%c is positioned at line n (the FREAD target variable is
   omitted for lack of space):

  { FREAD \%c, FSEEK /LINE \%c -1, FREAD \%c }  <-- Reads line n twice
  { FREAD \%c, FSEEK /LINE \%c +0, FREAD \%c }  <-- Reads lines n and n+1
  { FREAD \%c, FSEEK /LINE \%c +1, FREAD \%c }  <-- Reads lines n and n+2
  { FREAD \%c, FSEEK /LINE \%c -2, FREAD \%c }  <-- Reads lines n and n-1
  { FREAD \%c, FSEEK /LINE \%c -3, FREAD \%c }  <-- Reads lines n and n-2

   Another caution: Using FSEEK and FREAD /SIZE to repeatedly read the
   same disk block (e.g. when sampling a database record that is
   frequently updated) might not give you updated disk blocks due to the
   internal buffering and caching of the C library (this probably varies
   from one platform/compiler combination to another). If necessary you
   can force a fresh disk read with a close/open sequence:

  FCLOS \%c
  FOPEN \%c samefilename
  FSEEK \%c samespot
  FREAD /SIZE:howmanybytes \%c variable

1.22.3. FILE Command Examples

   To read the last 10 lines of a text file into an array:

  fopen /read \%c oofa.txt                ; Open the file
  if fail exit 1 Can't open oofa.txt      ; Always check for failure
  dcl \&a[10]                             ; Declare a 10-element array
  fcount /line \%c                        ; Count lines in the file
  fseek /line \%c \v(f_count)-10          ; Seek to 10 lines from the end
  if fail exit 1 Can't seek               ; Check for failure
  for \%i 1 10 1 { fread \%c \&a[\%i] }   ; Read the last 10 lines
  fclose \%c                              ; Close the file

   Note that blank lines show up as empty (undefined) array elements, for
   example if you give a "show array a" command at this point. This is
   normal. You can still use these elements; e.g.:

  for \%i 1 10 1 { echo \%i. \&a[\%i] }   ; Display the 10 lines

   Here is how to read the last line of a file (already open on channel
   \%c):

  fseek /line \%c last                    ; Seek directly to last line

   Alternatively:

  fseek /line \%c eof                     ; Seek to end of file
  fseek /line \%c -1                      ; Seek to beginning of last line

   Alternatively:

  fcount /line \%c                        ; Count the file's lines
  fseek /line \%c \v(f_count)-1           ; Seek to last line
  fread \%c                               ; Read it

   To read every other line from the file (using relative SEEK), skipping
   the first line:

  fopen /read \%c oofa.txt                ; Open the file
  while ( success ) {                     ; Loop through lines
      fseek /line \%c +1                  ; Skip a line
      if success fread \%c                ; Read & display a line
  }
  fclose \%c                              ; Close the file

   Here is how to read the lines of a file in reverse order:

  fopen /read \%c oofa.txt                ; Open
  if fail exit 1                          ; Check
  fseek /line \%c last                    ; Seek to last line
  while success {                         ; Loop
      fread \%c                           ; Read line
      fseek /line \%c -2                  ; Seek backwards two lines
  }
  fclose \%c                              ; Close the file

   The loop works because a relative SEEK outside the file fails.

   It is also possible to use block i/o to manage random-access files with
   fixed-length records (as long as they don't contain NUL characters).
   Suppose, for example, you have a file of "card image" records with
   fixed-field information about customers, such as:

  Name:     Columns  1-32  (column numbers are 1-based)
  Address:  Columns 33-72
  Balance:  Columns 73-80

   The records are indexed by customer number, starting with 0. There are
   no line terminators separating them. Therefore the record for customer
   number n starts at position nx 80 (\%n*80).

   Now suppose we received a payment from customer number 173 and want to
   update the balance:

  .\%n = 173                               ; Customer (record) number
  .\%a = 12.72                             ; Amount
  fopen /read /write \%c customer.db       ; Open the file
  if fail stop 1 OPEN FAILED: \f_errmsg()  ; Check
  fseek /byte \%c 80*\%n                   ; Seek to record
  fread /size:80 \%c r                     ; Read the record
  if fail stop 1 READ FAILED: \f_errmsg()  ; Check (IMPORTANT)
  .\%b := \fright(\m(r),8)                 ; Extract the balance
  .\%b := \ffpadd(\%b,\%a,2)               ; Add the new payment
  if fail stop 1 ARITHMETIC ERROR: \%b/\%a ; Catch bad records
  .r := {\fleft(\m(r),72)\flpad(\%b,8)}    ; Update the record
  fseek /byte \%c 80*\%n                   ; Reposition to same spot
  fwrite /size:80 \%c {\m(r)}              ; Replace the record
  if fail stop 1 WRITE FAILED: \f_errmsg() ; Check
  fclose \%c                               ; Close the file

   REMEMBER: Using FILE SEEK to move beyond the end of file can result in
   a file with holes when writing; when reading, an end-of-file error will
   occur -- be sure to check for it.

1.22.4. Channel Numbers

   C-Kermit's channel numbers are integers from 0 to some
   platform-dependent limit, such as 46 or 1985 (the value of \v(f_max)).
   This is the limit placed by the operating system on the number of files
   that may be opened by one process or user or job, minus the standard
   input, output, and error files, and minus the number of files reserved
   by C-Kermit for logs, OPEN READ and WRITE, and file transfer (and maybe
   some command files -- the \v(f_max) number can't be exact).

   Although you must include a variable in the FILE OPEN command, to which
   the channel number is assigned, you don't have to use a variable in the
   other FILE commands if you know what the number is -- you can just put
   the number. This saves you a few keystrokes when typing commands at the
   prompt:

  fopen \%c oofa.txt
  flist
  0. /usr/olga.oofa.txt (R) 0

   This tells the channel number is 0 (the number on the left is the
   channel file's channel number). Of course you can also find it by
   echoing the variable:

  echo \%c
  0

   Or with "fstatus \%c". Now you can type commands like:

  fread 0

   to read a line from the file. Obviously, however, using digits rather
   than a variable for the channel number would be poor practice in a
   script.

   If in commands like:

  fread \%c \%a

   you have trouble remembering which variable is which, note that the
   channel number is, indeed, a number. Anywhere C-Kermit accepts a number
   it can also accept an expression, so you can put parentheses around the
   channel number to remind you it's the channel number and not the
   variable into which data is to be read:

  fread (\%c) \%a

   Normally channel numbers are assigned sequentially as 0, 1, 2, ... up
   to the limit. However, once you start closing files, there can be holes
   in the sequence. New channels are assigned to fill in the holes. Thus
   you can't depend on channel numbers being in any particular sequence.

1.22.5. FILE Command Errors

   Each FILE command sets the variable \v(f_error) to one of the following
   values:

    0 = No error
   -1 = System error
   -2 = Attempt to read after end of file
   -3 = Channel not open
   -4 = Channel number out of range (negative or too large)
   -5 = Numeric argument (size, ...) out of range
   -6 = File not found
   -7 = Bad or missing filename
   -8 = Too many files are already open (FILE OPEN only)
   -9 = Forbidden operation (e.g. write to a read-only file)
  -10 = Access denied
  -11 = Illegal combination of OPEN modes (FILE OPEN only)
  -12 = Buffer overflow
  -13 = Current line number unknown (for relative line seeks)
  -14 through -98: Reserved.
  -99 = Requested operation not implemented in this version of C-Kermit
 -999 = Unknown error

   When \v(f_error) is -1, this means the FILE command failed because
   because of a system error, in which case you can examine the following
   variables:

  \v(errno)     = System error number.
  \v(errstring) = Error message corresponding to \v(errno).

   A special function is available for translating the \v(f_error) code to
   an error message string:

\f_errmsg([code])
  If the code is -1, returns error message of the most recent system
  error; otherwise if the code is a valid \v(f_error) value, the associated
  message is returned.  If the code is omitted, the status message
  corresponding to the current \v(f_error) value is returned.

   A FILE command that fails prints the appropriate error message
   automatically, except when the command is READ or SEEK and the error is
   -2 (end of file); in that case, the command still fails, but does not
   print a message. This allows constructions such as:

  fopen \%c oofa.txt
  while success { fread \%c }
  fclose \%c

   to work as expected, i.e. without an annoying message when the end of
   file is reached.

1.22.6. File I/O Variables

   The variables associated with the file i/o package are:

   \v(f_count)
          Result of the most recent FILE COUNT (FCOUNT) command.

   \v(f_error)
          Numeric error code of most recent FILE command (0 = no error).

   \v(f_max)
          Maximum number of files open simultaneously.

1.22.7. File I/O Functions

   Some of the FILE commands can also be issued as function calls, which
   makes script writing a bit more convenient, especially for C
   programmers. Also, several functions are provided that do not have
   command equivalents. Each of these functions takes a channel number as
   the first argument. These functions do not work for OPEN { READ, !READ,
   WRITE, !WRITE, and APPEND } files.

   \f_status(channel)
          Returns 0 if the channel is not open, otherwise a number between
          1 and 15 which is the sum of the OPEN modes:

  1 = /READ
  2 = /WRITE
  4 = /APPEND
  8 = /BINARY

   The remaining functions work only for open channels. Each of these
   functions can fail for the applicable reasons listed inSection
   1.22.5. For instructions on handling function errors, seeSection
   7.12.

   \f_pos(channel)
          Returns the file's current read/write pointer (0-based). There
          is no FILE command equivalent.

   \f_line(channel)
          Returns the file's current line number (0-based), if known,
          otherwise -1. There is no FILE command equivalent. The line
          number is known as long as no character or block i/o has been
          done on the channel.

   \f_handle(channel)
          Returns the "file handle" of the file. That is, it translates
          the portable C-Kermit channel number into a system-specific file
          handle or number that can be passed to other programs on the
          same platform. In UNIX this is a file descriptor. There is no
          FILE command equivalent.

   \f_eof(channel)
          Returns 1 if the read/write pointer of the file on the given
          channel is at the end of the file, 0 otherwise. Convenient in
          WHILE statements, e.g.:

    while not \f_eof(\%c) { fread \%c }

   \f_getchar(channel)
          Equivalent to FREAD /CHAR. Returns the character actually read.
          If \f_getchar() does not fail but the return value is empty,
          this means a NULL character was read.

   \f_getline(channel)
          Equivalent to FREAD /LINE. Returns the line actually read, but
          with the line terminator stripped. If \f_getline() does not fail
          but the return value is empty, this normally means an empty line
          was read.

   \f_getblock(channel,n)
          Equivalent to FREAD /SIZE:n. Returns the block of characters
          actually read. If the returned block is smaller than n, it
          indicates either that the end of file was reached or a NUL
          character is in the block.

   \f_putchar(channel,c)
          Equivalent to FWRITE /CHARACTER. Writes the character c. If c
          contains more than one character, only the first is written. If
          c is empty a NUL is written. Returns the number of characters
          written on success, or a negative error code upon failure.

   \f_putline(channel,string)
          Equivalent to FWRITE /LINE. Writes the string and adds the
          appropriate line termination character or sequence. If the
          string is empty or omitted, an empty line is written. Returns
          the number of characters written on success, or a negative error
          code upon failure.

   \f_putblock(channel,string)
          Equivalent to FWRITE /STRING. Writes the string as given. If the
          string is empty or omitted, nothing is written. Returns the
          number of characters written on success, or a negative error
          code upon failure.

1.22.8. File I/O Function Examples

  fopen /read \%c oofa.txt            ; Open our favorite file for reading
  if failure exit 1                   ; Check that it's open
  while not \f_eof(\%c) {             ; Loop until EOF
      .line := \f_getline(\%c)        ; Get a line
      if success echo {\m(line)}      ; Echo it
  }
  if not \f_eof(\%c) {                ; Check reason for loop exit
      exit 1 File Error: \f_errmsg()  ; If not EOF say so.
  }

  frewind \%c                         ; Rewind the file
  while not \f_eof(\%c) {             ; Same thing but with block i/o
      .block := \f_getblock(\%c,256)  ; (much faster than line i/o)
      if success xecho {\m(block)}
  }

  frewind \%c                         ; Rewind again
  while not \f_eof(\%c) {             ; Same deal but with character i/o
      .c := \f_getchar(\%c)           ; (much slower than line i/o)
      if success xecho {\m(c)}
  }
  close \%c

   To close all open files (equivalent to FCLOSE ALL):

  for \%i 0 \v(f_max)-1 1 {
      if \f_status(\%i) fclose \%i
  }

1.23. The EXEC Command

   The EXEC command is available only in UNIX.

   EXEC [ /REDIRECT ] command [ arg1 [ arg2 [ ... ] ]
          Runs the given command with the arguments in such a way that the
          command replaces C-Kermit in memory, and C-Kermit ceases to
          execute. EXEC is like RUN, except instead of returning to
          C-Kermit when finished, the command returns to whatever process
          invoked Kermit.

   In the normal case, no files are closed, so the EXEC'd command inherits
   the open files, read/write pointers, working directory, process ID,
   user ID (unless command is SUID), group ID (unless command is SGID),
   groups, etc. (In UNIX, the EXEC command is simply a front end for
   execvp().)

   If the /REDIRECT switch is included, then if a connection is open (SET
   LINE or SET HOST), it becomes the standard input and output of the
   EXEC'd program. If no connection is open, the /REDIRECT switch has no
   effect. For example to use C-Kermit for PPP dialing in Linux:

  set modem type usr          ; Specify the kind of modem you have
  set line /dev/ttyS1         ; Specify the device it's connected to
  set speed 57600             ; and the speed
  set flow rts/cts            ; and flow control.
  set dial retries 100        ; Try the dial sequence up to 100 times.
  dial {{9-212-555-1212}{9-212-555-1213}{9-212-555-1214}{9-212-555-1215}}
  if fail exit 1
  for \%i 1 16 1 {            ; Try up to 16 times to get login prompt
      input 10 Login:         ; Wait 10 sec for it to appear
      if success break        ; Got it - proceed...
      output \13              ; Send a carriage return and try again
  }
  if ( > \%i 16 ) stop 1 NO LOGIN PROMPT
  lineout \(myuserid)         ; Send user ID
  input 30 assword:           ; Wait for Password prompt
  if fail stop 1 NO PASSWORD PROMPT
  lineout \m(mypassword)      ; Send the password.
  exec /redirect pppd         ; Replace ourselves with pppd.

   In this example we assume that the script has already set up the
   myuserid and mypassword variables -- normally the password should be
   prompted for, rather than stored on disk. Notice the advantages over
   the well-known "chat script":
     * You don't have to control the modem itself with AT commands;
       Kermit's DIAL command does this for you.
     * You can have Kermit automatically redial as many times as you want
       until it gets a connection (if this is legal in your country).
     * You can have Kermit fetch the number or numbers from a dialing
       directory.
     * You can have Kermit cycle through a list of phone numbers (this is
       new in C-Kermit 7.0; see Section 2.1.16) without having to
       enter the numbers in a dialing directory.
     * Dialing is location-independent; you can use the same script to
       dial from different areas or countries.
     * Once the connection is made, the full power of Kermit's script
       language is available to manage the dialog with the terminal server
       or other device that answers the phone call.

   NOTE: PPP and SLIP dialing are not available in Windows 95/98/NT/2000,
   whose APIs do not provide a method for an application to hand over a
   connection to the PPP or SLIP driver.

1.24. Getting Keyword Lists with '?'

   Suppose you type "te" at the C-Kermit> 6.0 prompt and then Esc or Tab
   to request keyword completion. Kermit beeps, indicating that more than
   one command starts with "te". But if you type '?' to see what they are,
   Kermit shows only "telnet". So why the beep? Because of invisible
   keywords like "telopt", "terminal", and "text". Lots of keywords are
   invisible because they are either synonyms for other keywords or else
   esoteric options to be used only in special circumstances, so we don't
   want them cluttering up the menus.

   But then there is no way for you to discover them. So in C-Kermit 7.0,
   if you type '?' AFTER the beginning of a keyword field, then invisible
   keywords are shown too:

  C-Kermit> te<Esc><BEEP>
  C-Kermit> te? Command, one of the following:
   telnet    telopt    terminal  text
  C-Kermit>te

   But if '?' is typed at the beginning of a field, only visible keywords
   are shown, as before (so, in this example, if '?' is typed at the
   C-Kermit> prompt, "telnet" is the only command shown that starts with
   "te").

2. MAKING AND USING CONNECTIONS The SET LINE, SET HOST, and SET PORT (a
synonym for SET LINE) commands have new synonyms, in which the word SET is
replaced by the word OPEN: OPEN LINE, etc. There is no new functionality
here, but OPEN is a better verb, since SET generally takes no action, whereas
these commands actually try to open a connection. Furthermore, there is the
symmetry with CLOSE. 2.0. SET LINE and SET HOST Command SwitchesThe SET LINE
(SET PORT) and SET HOST commands now allow switches before the device or host
name, in most cases, and under certain circumstances, also at the end. The
new syntax is backwards compatible with the previous syntax; thus SET LINE,
SET PORT, and SET HOST commands in command files written for C-Kermit 6.0 or
earlier still work. The expanded syntax is:

{ OPEN, SET } { LINE, PORT, HOST } [ switches ] device-or-address [ switches
]

The first group of switches is:

   /NETWORK-TYPE:{TCP/IP,X.25,PIPE,PTY...}
          When more than one network type is available, this lets you
          specify the type of network to use for this connection without
          affecting your global SET NETWORK TYPE. See Section 2.7
          about pipes and ptys.

   /USERID:[string]
          This switch is equivalent to SET LOGIN USERID. If a string is
          given, it sent to host during Telnet negotiations; if this
          switch is given but the string is omitted, no user ID is sent to
          the host. If this switch is not given, your current LOGIN USERID
          (\v(userid) value), if any, is sent. Unlike most other switches,
          this one is "sticky", since the value must persist throughout
          the session in case the server requests the ID string at a later
          time.

   /CONNECT
          Enter CONNECT mode immediately and automatically after the
          device or connection is open. On serial devices, however, when
          CARRIER-WATCH is not OFF, wait up to 1 second for the Carrier
          Detect signal to appear before trying to connect, to give the
          device time to react DTR, which might have been off prior to
          opening the device.

   /SERVER
          Enter server mode immediately and automatically after the device
          or connection is open. Treatment of carrier is the same as for
          /CONNECT.

   /WAIT
   /NOWAIT
          For Telnet connections only: Like SET TELNET WAIT { ON, OFF },
          but applies only to this connection, and in fact applies only
          when OPENing this connection (which is usually the only place it
          matters). Typically you would use TELNET /NOWAIT to make a
          connection to a misbehaving Telnet server that does not reply to
          negotiations as required by the Telnet protocol definition.

   Note: /CONNECT and /SERVER switches are not available in the RLOGIN and
   TELNET commands, since these commands already include an implicit
   /CONNECT and preclude automatic entry into server mode.

   The /CONNECT and /SERVER switches are especially useful with "set host
   *" connections. For example, suppose you want to start a Kermit server
   on socket 3000 of your TCP host. Normally you would have to give the
   command:

  set host * 3000

   and then wait for a connection to come in, and only then could you give
   the SERVER command (or else define a macro to do this, and then execute
   the macro). Now you can do it in one step:

  set host /server * 3000

   This tells C-Kermit to wait for the connection and then enter server
   mode once it comes in, no matter how long it takes. Similarly, "set
   host /conn *" can be used to wait for a "chat" connection to come in.

   Another set of switches is available in VMS only, for use only with SET
   LINE:

   /SHARE
          Allows the SET LINE device to be opened in shared mode. Normally
          it makes no sense to open a serial device in shared mode, but
          it's necessary when C-Kermit is running in an environment such
          as DECIntact, that opens your job's controlling terminal in such
          a way that C-Kermit can't open it too, unless it enables SHARE
          privilege. Note: SHARE privilege is required.

   /NOSHARE
          Requires that the SET LINE device not be in use by any other
          process in order for it to be successfully opened by C-Kermit.
          If neither /SHARE nor /NOSHARE is specified, /NOSHARE is used.

   The second group of switches is:

   /NO-TELNET-INIT
          Do not send initial Telnet negotiations even if this is a Telnet
          port.

   /RAW-SOCKET
          This is a connection to a raw TCP socket ( Section 2.3.5).

   /RLOGIN
          Use Rlogin protocol even if this is not an Rlogin port.

   /TELNET
          Send initial Telnet negotiations even if this is not a Telnet
          port.

   As of C-Kermit 7.0 and K95 1.1.19, the TELNET command includes an
   implicit /TELNET switch. So if you TELNET to a non-TELNET port, Kermit
   sends initial Telnet negotiations. This makes sense, since that's what
   "telnet" means.

   If you want to make a connection to a non-Telnet port without sending
   initial Telnet negotiations, use:

  set host [ /connect ] name-or-address port

   or:

  telnet name-or-address port /no-telnet-init

   Additional switches might be added in the future; type "set host ?" or
   "set line ?" to see a current list.

2.1. Dialing

   Automatic redialing is illegal or restricted in many countries, so
   until C-Kermit 7.0, it was disabled by default, i.e. until a SET DIAL
   RETRIES command was given. In C-Kermit 7.0, if no SET DIAL RETRIES
   command has been given, a default is picked dynamically at DIAL time
   based on the calling country code, if known. At this writing, the only
   country code known to have no restrictions on automatic redialing is 1.
   So in this case a limit of 10 is chosen; otherwise 1. If you have not
   given an explicit SET DIAL RETRIES command, SHOW DIAL shows the value
   as "(auto)", and then the value actually used is shown when you give
   the DIAL command.

   As of C-Kermit 7.0, automatic redialing is automatically canceled if
   the call could not be placed because no dialtone was detected.

2.1.1. The Dial Result Message

   If DIAL DISPLAY is not ON, the "Call complete" message now shows the
   modem's call result message, for example:

  Dialing: ...
  Call complete: "CONNECT 31200/ARQ/V32/LAPM/V42BIS"

   The exact format and contents of this message, of course, depends on
   the make, model, and configuration of your modem, so use your modem
   manual to interpret it. The call result message is also available in
   C-Kermit's \v(dialresult) variable.

  C-Kermit> echo \v(dialresult)
  CONNECT 31200/ARQ/V32/LAPM/V42BIS
  C-Kermit> echo Speed = \fword(\v(dialresult),2)
  Speed = 31200
  C-Kermit>

   Suppose your modem reports the modulation speed as shown above and you
   want to ensure your call is completed at (say) 24000 bps or more. You
   can use a little macro to do the job:

define HSDIAL {                ; High Speed DIAL
    local \%s
    if < \v(argc) 1 if not def \v(dialnumber) end 1 Usage: \%0 number
    set dial retries 100
    set dial interval 1
    while true {
        dial \%*
        if fail end 1 DIAL failed.
        asg \%s \fword(\v(dialresult),2)
        if def \%s if numeric \%s if not < \%s 24000 break
    }
}

   (See Section 7.5 about the \%* variable.)

2.1.2. Long-Distance Dialing Changes

   Due to the glut of cell phones, pagers, fax machines, ISPs, etc, area
   codes and dialing rules are changing all the time. In the North
   American Numbering Plan (NANP) countries (USA, Canada, etc), area codes
   are split or overlayed with increasing frequency, and 10- and 11-digit
   dialing is gradually becoming the norm for local calls. Changes are
   occurring In Europe, too, partly for these reasons and partly because
   of some new EC rules.

   In France, effective 18 October 1996, all calls, even local ones, must
   be dialed with an area code. French area codes are presently 1-digit
   numbers, 1-6, and the long-distance dialing prefix is 0. All calls
   within France are considered long distance and begin with 01, 02, ...,
   06.

   Effective 1 May 1997, all calls within the US state of Maryland, even
   local ones, must be dialed with an area code but without the
   long-distance prefix -- this is the now widely-known North American
   phenomenon of "ten digit dialing". The same is happening elsewhere --
   many cities in Florida adopted 10-digit dialing in 1998.

   In Italy beginning 19 June 1998, all calls to fixed (as opposed to
   mobile) numbers must be prefixed by 0. When calling into Italy from
   outside, the 0 must follow the country code (39). Calls to cell phones,
   however, must be placed without the 0. Then on 29 December 2000, the 0
   will become a 4 (for calling fixed numbers) and a prefix of 3 must used
   for calling mobile phones. More info at:
   http://www.telecomitalia.it/npnn/.

   In Spain, effective 4 April 1998, with hard cutover on 18 July 1998,
   all calls within the country must be dialed with 9 digits, and all
   calls from outside Spain must also be dialed with 9 digits (after the
   country code, 34). The new 9-digit numbers all begin with "9". More
   info at:http://www.telefonica.es/cambiodenumeracion/

   Several new dialing features and commands have been added in version
   6.1 and 7.0 to address these changes.

   C-Kermit 6.0 and Kermit 95 1.1.11 and earlier handle the French
   situation via a reasonable subterfuge (setting the local area code to a
   nonexistent one), but did not handle "ten-digit dialing" well at all;
   the recommended technique was to change the long-distance dialing
   prefix to nothing, but this defeated Kermit's "list numbers for one
   name" feature when the numbers were in different locations. For
   example:

  set dial ld-prefix
  dial onlineservice

   where "onlineservice" is a dialing directory entry name corresponding
   to entries that are in (say) Maryland as well as other states, would
   not correctly dial the numbers not in Maryland.

   A new command lets you specify a list of area codes to be considered
   local, except that the area code must be dialed:

   SET DIAL LC-AREA-CODES [ areacode [ areacode [ areacode [ ... ] ] ] ]
          The list may include up to 32 area codes. If a number is called
          whose area code is in this list, it is dialed WITHOUT the
          long-distance prefix, but WITH the area code.

   So in Maryland, which (last time we looked) has two area codes, 410 and
   301, the setup would be:

  SET DIAL LC-AREA-CODES 410 301

   Example:

  SET DIAL LD-PREFIX 1
  SET DIAL AREA-CODE 301
  SET DIAL LC-AREA-CODES 410 301 <-- Area codes in 10-digit dialing region
  DIAL +1 (301) 765-4321         <-- Dials 3017654321  (local with area code)
  DIAL +1 (410) 765-4321         <-- Dials 4107654321  (local with area code)
  DIAL +1 (212) 765-4321         <-- Dials 12127654321 (long distance)

   The SET DIAL LC-AREA-CODES command does not replace the SET DIAL
   AREA-CODE command. The latter specifies the area code you are dialing
   from. If the called number is in the same area code, then the area code
   is dialed if it is also in the LC-AREA-CODES list, and it is not dialed
   otherwise. So if "301" had not appeared in the LC-AREA-CODES list in
   the previous example:

  SET DIAL LD-PREFIX 1
  SET DIAL AREA-CODE 301
  SET DIAL LC-AREA-CODES 410     <-- Area codes in 10-digit dialing region
  DIAL +1 (301) 765-4321         <-- Dials 7654321     (local)
  DIAL +1 (410) 765-4321         <-- Dials 4107654321  (local with area code)
  DIAL +1 (212) 765-4321         <-- Dials 12127654321 (long distance)

   The new Kermit versions also add a Local Call Prefix and Local Call
   Suffix, in case you have any need for it. These are added to the
   beginning and of local phone numbers (i.e. numbers that are not
   long-distance or international). Examples:

  SET DIAL LD-PREFIX 1
  SET DIAL LC-PREFIX 9
  SET DIAL LC-SUFFIX *
  SET DIAL LC-AREA-CODES 410     <-- Area codes in 10-digit dialing region
  SET DIAL AREA-CODE 301
  DIAL +1 (301) 765-4321         <-- Dials 97654321*     (local)
  DIAL +1 (410) 765-4321         <-- Dials 94107654321*  (local with area code)
  DIAL +1 (212) 765-4321         <-- Dials 12127654321   (long distance)

2.1.3. Forcing Long-Distance Dialing

   Suppose a number is in your country and area, but for some reason you
   need to dial it long-distance anyway (as is always the case in France).
   There have always been various ways to handle this:

    1. Temporarily set your area code to a different (or nonexistent or
       impossible) one (but this required knowledge of which area codes
       were nonexistent or impossible in each country).
    2. Dial the number literally instead of using the portable format, but
       this defeats the purpose of the portable dialing directory.

   Now there is also a new command that, very simply, can force
   long-distance dialing:

   SET DIAL FORCE-LONG-DISTANCE { ON, OFF }
          If a call is placed to a portable phone number within the same
          country code as the calling number, it is dialed with the
          long-distance prefix and the area code if FORCE-LONG-DISTANCE is
          ON. If OFF, the regular rules and procedures apply.

   Example (France):

  SET DIAL COUNTRY-CODE 33
  SET DIAL AREA-CODE 6
  SET DIAL FORCE-LONG-DISTANCE ON

   (In fact, SET DIAL COUNTRY-CODE 33 automatically sets DIAL
   FORCE-LONG-DISTANCE ON...)

   Example (USA, for a certain type of reverse-charge calling in which the
   called number must always be fully specified):

  SET DIAL PREFIX 18002666328$     ; 1-800-COLLECT
  SET DIAL COUNTRY-CODE 1
  SET DIAL AREA-CODE 212
  SET DIAL FORCE-LONG-DISTANCE ON

   Example (Toronto, where calls to exchange 976 within area code 416 must
   be dialed as long distance, even when you are dialing from area code
   416):

  SET DIAL COUNTRY-CODE 1
  SET DIAL AREA-CODE 416
  SET DIAL FORCE-LONG-DISTANCE ON
  DIAL +1 (416) 976-xxxx

   If dialing methods were consistent and sensible, of course it would be
   possible to always dial every domestic call as if it were long
   distance. But in many locations this doesn't work or if it does, it
   costs extra. The following macro can be used for dialing any given
   number with forced long-distance format:

  define LDIAL {
      local \%x
      set dial force-long-distance on
      dial \%*
      asg \%x \v(success)
      set dial force-long-distance off
      end \%x
  }

   (See Section 7.5 about the \%* variable.)

2.1.4. Exchange-Specific Dialing Decisions

   This applies mainly to the North American Numbering Plan (NANP). Refer
   to the section "Alternative notations" inUsing C-Kermit 2nd
   Edition, pages 106-107, and the story about Toronto on page 110. Using
   the new LC-AREA-CODES list, we can address the problem by treating the
   exchange as part of the area code:

  SET DIAL LD-PREFIX 1
  SET DIAL AREA-CODE 416
  SET DIAL LC-AREA-CODES 905276
  DIAL +1 416 765 4321               <-- 7654321      (local)
  DIAL +1 905 276 4321               <-- 9052764321   (local with area code)
  DIAL +1 905 528 4321               <-- 19055284321  (long distance)

   The same technique can be used in Massachusetts (story at top of page
   111) and in any other place where dialing to some exchanges within a
   particular area code is local, but to others in the same area code is
   long distance.

2.1.5. Cautions about Cheapest-First Dialing

   Kermit does not maintain a knowledge base of telephony information; it
   only provides the tools to let you enter a phone number in a standard
   format and dial it correctly from any location in most cases.

   In particular, Kermit does not differentiate the charging method from
   the dialing method. If a call that is DIALED as long-distance (e.g.
   from 212 to 718 in country code 1) is not CHARGED as long distance, we
   have no way of knowing that without keeping a matrix of charging
   information for every area-code combination within every country, and
   any such matrix would be obsolete five minutes after it was
   constructed. Thus, "cheapest-first" sorting is only as reliable as our
   assumption that the charging method follows the dialing method. A good
   illustration would be certain online services that have toll-free
   dialup numbers which they charge you a premium (in your online service
   bill) for using.

2.1.6. Blind Dialing (Dialing with No Dialtone)

   C-Kermit's init string for Hayes-like modems generally includes an X4
   command to enable as many result codes as possible, so that Kermit can
   react appropriately to different failure reasons. One of the result
   codes that X4 enables is "NO DIALTONE". A perhaps not obvious side
   effect of enabling this result code that the modem must hear dialtone
   before it will dial.

   It is becoming increasingly necessary to force a modem to dial even
   though it does not hear a dialtone on the phone line; for example, with
   PBXs that have strange dialtones, or with phone systems in different
   countries, or with ISDN phones, etc. This is called "blind dialing".

   C-Kermit 7.0 has two new commands to cope with this situation:

   SET DIAL IGNORE-DIALTONE { ON, OFF }
          OFF (the default) means to tell the modem to wait for dialtone
          before dialing. ON means to enable "blind dialing", i.e. tell
          the modem NOT to wait for dialtone before dialing. Generally
          this is accomplished by sending ATX3 to the modem just prior to
          dialing. SET MODEM TYPE xxx and then SHOW MODEM displays
          Kermit's built-in "ignore dialtone" command.

   SET DIAL COMMAND IGNORE-DIALTONE text
          This lets you change the built-in ignore-dialtone command (such
          as ATX3) to whatever you choose, in case the built-in one does
          not work, or another command works better.

   Notes:
    1. The ignore-dialtone command is not sent unless SET DIAL
       IGNORE-DIALTONE is ON.
    2. The ATX3 command generally disables not only NO DIALTONE, but also
       BUSY. So this will prevent Kermit from detecting when the line is
       busy. This is a property of the modem, not of Kermit.

2.1.7. Trimming the Dialing Dialog

   The command:

  SET MODEM COMMAND action [ command ]

   is used to override Kermit's built-in modem commands for each action,
   for each kind of modem in its internal database. If you include a
   command, this is used instead of the built-in one. If you omit the
   command, this restores the original built-in command.

   If you want to omit the command altogether, so Kermit doesn't send the
   command at all, or wait for a response, use:

  SET MODEM COMMAND action {}

   That is, specify a pair of empty braces as the command, for example:

  SET MODEM COMMAND ERROR-CORRECTION ON {}

2.1.8. Controlling the Dialing Speed

   The rate at which characters are sent to the modem during dialing is
   normally controlled by the built-in modem database. You might want to
   override this if Kermit seems to be dialing too slowly, or it is
   sending characters to the modem faster than the modem handle them. A
   new command was added for this in C-Kermit 7.0:

   SET DIAL PACING number
          Specifies the number of milliseconds (thousandths of seconds) to
          pause between each character when sending commands to the modem
          during DIAL or ANSWER command execution. 0 means no pause at
          all, -1 (the default) or any other negative number means to use
          the value from the database. Any number greater than 0 is the
          number of milliseconds to pause.

   HINT: You might also need to control the rate at which the modem
   generates Touch Tones during dialing, for example when sending a
   numeric page. There are two ways to do this. One way is to insert pause
   characters into the dialing string. For modems that use the AT command
   set, the pause character is comma (,) and causes a 2-second pause. On
   most modems, you can use the S8 register to change the pause interval
   caused by comma in the dialing string. The other way is to set your
   modem's tone generation interval, if it has a command for that. Most
   AT-command-set modems use S11 for this; the value is in milliseconds.
   For example on USR modems:

  ATS11=200

   selects an interval of 200 milliseconds to separate each dialing tone.

   Hint: To add S-Register settings or other commands to your dialing
   procedure, use the new SET MODEM COMMAND PREDIAL-INIT command
   ( Section 2.2.2).

2.1.9. Pretesting Phone Number Conversions

   The LOOKUP command now accepts telephone numbers as well as
   directory-entry names, for example:

  LOOKUP +1 (212) 7654321

   When given a phone number, LOOKUP prints the result of converting the
   phone number for dialing under the current dialing rules. For example,
   if my country code is 1 and my area code is 212, and I am dialing out
   from a PBX whose outside-line prefix is "93,":

  C-Kermit> lookup +1 (212) 7654321
  +1 (212) 7654321 => 93,7654321
  C-Kermit>

   You can also use the \fdialconvert(phone-number) function (Section
   2.1.11) to do this programmatically:

  C-Kermit> echo "\fdialconvert(+1 (212) 7654321)"
  "93,7654321"
  C-Kermit>

   So the new LOOKUP behaves as follows:

   LOOKUP portable-format-phone-number
          Displays how the number would actually be dialed Sets FAILURE if
          there was a conversion error, otherwise SUCCESS.

   LOOKUP literal-format-phone-number
          Displays the same literal-format-phone-number Always sets
          SUCCESS.

   LOOKUP dialing-directory-name
          Displays all matching entries and converts portable phone
          numbers. Sets SUCCESS if at least one entry was found, otherwise
          FAILURE.

   LOOKUP =anything
          Displays "=anything" and sets SUCCESS.

   There is, at present, no programmatic way to fetch numbers from the
   dialing directory. This will be considered for a future release.

2.1.10. Greater Control over Partial Dialing

   The following rules now apply to partial dialing:

     * Phone number transformations based on country and area code,
       application of prefixes, etc, are performed only on the first
       PDIAL.
     * Each PDIAL argument is looked up in the dialing directory, so it is
       possible have directory entries for pieces of phone numbers or
       other information.
     * Suffixes are not applied automatically, since there is no way for
       C-Kermit to know in which PDIAL segment you want them to be
       applied.

   However, the suffix that *would* have been applied, based on the
   dialing rules that were invoked when processing the first PDIAL
   command, is stored in the variable:

  \v(dialsuffix)

   which you can include in any subsequent PDIAL or DIAL commands.

   Example:

  pdial {\m(my_long_distance_pager_number_part_1)}
  pdial {\m(my_long_distance_pager_number_part_2)}
  pdial {\v(dialsuffix)}
  pdial {\m(my_long_distance_pager_number_part_3)}
  pdial {@\m(numeric_pager_code)#}

2.1.11. New DIAL-related Variables and Functions

   \fdialconvert(s)
          s is a phone number in either literal or portable format (not a
          dialing directory entry name). The function returns the dial
          string that would actually be used by the DIAL command when
          dialing from the current location, after processing country
          code, area code, and other SET DIAL values, and should be the
          same as the result of LOOKUP when given a telephone number.

   \v(dialsuffix)
          Contains the suffix, if any, that was applied in the most recent
          DIAL command, or the suffix that would have been applied in the
          most recent PDIAL command. Use this variable to send the dial
          suffix at any desired point in a PDIAL sequence.

   \v(dialtype)
          A number indicating the type of call that was most recently
          placed. Can be used after a normal DIAL command, or after the
          first PDIAL command in a PDIAL sequence. Values are:

  -2: Unknown because TAPI handled the phone number translation.
  -1: Unknown because some kind of error occurred.
   0: Internal within PBX.
   1: Toll-free.
   2: Local within calling area.
   3: Unknown (e.g. because a literal-format phone number was given).
   4: Long distance within country.
   5: International

   \v(dialcount)
          The current value of the DIAL retry counter, for use in a DIAL
          macro ( Section 2.1.13).

   \v(d$px)
          PBX Exchange (see Section 2.1.12).

   Other dial-related variables, already documented inUsing C-Kermit
   (or other sections of this document, e.g. Section 2.1.1), include
   \v(dialnumber), \v(dialstatus), etc. A convenient way to display all of
   them is:

  show variable dial  ; hint: abbreviate "sho var dial"

   This shows the values of all the variables whose names start with
   "dial". Also "show variable d$" (to show the \v(d$...) variables).

2.1.12. Increased Flexibility of PBX Dialing

   Refer toUsing C-Kermit, 2nd Edition, pages 107-108. Recall that
   three commands are needed to configure C-Kermit for dialing from a PBX:

  SET DIAL PBX-EXCHANGE number
  SET DIAL PBX-INSIDE-PREFIX number
  SET DIAL PBX-OUTSIDE-PREFIX number

   Unfortunately, this model does not accommodate PBXs that have more than
   one exchange. For example our PBX at Columbia University (which must
   handle more than 10,000 phones) has 853-xxxx and 854-xxxx exchanges.

   Beginning in C-Kermit 7.0, the SET DIAL PBX-EXCHANGE command accepts a
   list of exchanges, e.g.:

  SET DIAL PBX-EXCHANGE 853 854

   (multiple exchanges are separated by spaces, not commas).

   So now when dialing a portable-format number that has the same country
   and area codes as those of your dialing location, C-Kermit compares the
   exchange of the dialed number with each number in the PBX Exchange list
   (rather than with a single PBX Exchange number, as it did formerly) to
   determine whether this is an internal PBX number or an external call.
   If it is an external call, then the PBX Outside Prefix is applied, and
   then the normal dialing rules for local or long-distance calls.

   If it is an inside call, the exchange is replaced by the PBX Inside
   Prefix. But if the PBX has more than one exchange, a single fixed PBX
   Inside Prefix is probably not sufficient. For example, at Columbia
   University, we must dial 3-xxxx for an internal call to 853-xxxx, but
   4-xxxx for a call to 854-xxxx. That is, the inside prefix is the final
   digit of the exchange we are dialing. For this reason, C-Kermit 7.0
   provides a method to determine the inside prefix dynamically at dialing
   time, consisting of a new variable and new syntax for the SET DIAL
   PBX-INSIDE-PREFIX command:

   \v(d$px)
          This variable contains the exchange that was matched when a PBX
          internal call was detected. For example, if the PBX exchange
          list is "853 854" and a call is placed to +1 (212) 854-9999,
          \v(d$px) is set to 854.

   SET DIAL PBX-INSIDE-PREFIX \fxxx(...)
          If the PBX Inside Prefix is defined to be a function, its
          evaluation is deferred until dialing time. Normally, this would
          be a string function having \v(d$px) as an operand. Of course,
          you can still specify a constant string, as before.

   So given the following setup:

  SET DIAL COUNTRY-CODE 1
  SET DIAL AREA-CODE 212
  SET DIAL PBX-OUTSIDE-PREFIX 93,
  SET DIAL PBX-EXCHANGE 853 854
  SET DIAL PBX-INSIDE-PREFIX \fright(\v(d$px),1)

   The following numbers give the results indicated:

 Number                   Result
  +1 (212) 854-9876        4-9876
  +1 (212) 853-1234        3-1234
  +1 (212) 765-4321        93,765-4321
  +1 (333) 765-4321        93,1333765-4321

   Furthermore, the K_PBX_XCH environment variable may now be set to a
   list of exchanges to automatically initialize C-Kermit's PBX exchange
   list, for example (in UNIX ksh or bash):

  export K_PBX_XCH="853 854"

   (Quotes required because of the space.) Of course, this variable can
   also be set to a single exchange, as before:

  export K_PBX_XCH=853

2.1.13. The DIAL macro - Last-Minute Phone Number Conversions

   After a DIAL or LOOKUP command is given, a list of phone numbers is
   assembled from the dialing directory (if any), with all
   location-dependent conversion rules applied as described in Chapter 5
   ofUsing C-Kermit.

   However, additional conversions might still be required at the last
   minute based on local or ephemeral conditions. So that you can have the
   final word on the exact format of the dial string, C-Kermit 7.0 lets
   you pass the converted string through a macro of your own design for
   final processing before dialing. The relevant command is:

   SET DIAL MACRO [ name ]
          Specifies the name of a macro to be run on each phone number
          after all built-in conversions have been applied, just before
          the number is dialed. If no name is given, no macro is run. The
          phone number, as it would have been dialed if there were no dial
          macro, is passed to the macro.

   The dial macro can do anything at all (except start a file transfer).
   However, the normal use for the macro would be to modify the phone
   number. For this reason the phone number is passed to the macro as
   argument number 1 (\%1). To cause a modified number to be dialed, the
   macro should terminate with a RETURN statement specifying a return
   value. To leave the number alone, the macro should simply end. Example:

  define xxx return 10108889999$\%1
  set dial macro xxx
  dial xyzcorp

   This defines a DIAL MACRO called xxx, which puts an access code on the
   front of the number. Another example might be:

  def xxx if equal "\v(modem)" "hayes-1200" return \freplace(\%1,$,{,,,,,})
  set dial macro xxx
  dial xyzcorp

   which replaces any dollar-sign in the dial string by a series of five
   commas, e.g. because this particular modem does not support the "wait
   for bong" feature (remember that commas that are to be included
   literally in function arguments must be enclosed in braces to
   distinguish them from the commas that separate the arguments) and when
   the IF condition is not satisfied, the macro does not return a value,
   and so the number is not modified. Then when a DIAL command is given
   referencing a dialing directory entry, "xyzcorp". The macro is
   automatically applied to each matching number.

   Numerous dial-, modem-, communications-, and time-related variables are
   available for decision making your dial macro. Type SHOW VARIABLES for
   a list. Of particular interest is the \v(dialcount) variable, which
   tells how many times the DIAL command gone through its retry loop: 1 on
   the first try, 2 on the second, 3 on the third, and so on, and the
   \v(dialresult) and \v(dialstatus) variables.

   Here are some other applications for the DIAL MACRO (from users):

     * Phone numbers in the dialing directory are formatted with '-' for
       readability, but some modems don't like the hyphens, so the DIAL
       macro is used to remove them before dialing; e.g 0090-123-456-78-99
       becomes 00901234567899: "def xxx return \freplace(\%1,-)".
     * To set some specific modem (or other) options depending on the
       called customer or telephone number.
     * Choosing the most appropriate provider based on (e.g.) time of day,
       or cycling through a list of providers in case some providers might
       be busy.

   To illustrate the final item, suppose you have a choice among many
   phone service providers; the provider is chosen by dialing an access
   code before the number. Different providers might be better (e.g.
   cheaper) for certain times of day or days of the week, or for dialing
   certain locations; you can use the DIAL macro to add the access for the
   most desirable provider.

   Similarly, when the same number might be reached through multiple
   providers, it's possible that one provider might not be able to
   complete the call, but another one can. In that case, you can use the
   DIAL macro to switch providers each time through the DIAL loop --
   that's where the \v(dialcount) variable comes in handy.

   The following command can be used to debug the DIAL macro:

   SET DIAL TEST { ON, OFF }
          Normally OFF, so the DIAL command actually dials. When ON, the
          DIAL command performs all lookups and number conversions, and
          then goes through the number list and retry loop, but instead of
          actually dialing, lists the numbers it would have called if none
          of the DIAL attempts succeeded (or more precisely, every number
          was always busy).

2.1.14. Automatic Tone/Pulse Dialing Selection

   SET DIAL METHOD { AUTO, DEFAULT, PULSE, TONE }
          Chooses the dialing method for subsequent calls.

   Prior to version 7.0, C-Kermit's DIAL METHOD was DEFAULT by default,
   meaning it does not specify a dialing method to the modem, but relies
   on the modem to have an appropriate default dialing method set. So, for
   example, when using Hayes compatible modems, the dial string would be
   something like ATD7654321, rather than ATDT7654321 or ATDP7654321.

   In C-Kermit 7.0 and K95 1.1.19, the dial method can be set from the
   environment variable:

  K_DIAL_METHOD

   when Kermit starts. The values can be TONE, PULSE, or DEFAULT, e.g.
   (UNIX):

  set K_DIAL_METHOD=TONE; export K_DIAL_METHOD

   In the absence of a K_DIAL_METHOD definition, the new default SET DIAL
   METHOD is AUTO rather than DEFAULT. When DIAL METHOD is AUTO and the
   local country code is known, then if tone dialing is universally
   available in the corresponding area, tone dialing is used; if dialing
   from a location where pulse dialing is mandatory, pulse dialing is
   used.

   The "tone country" and "pulse country" lists are preloaded according to
   our knowledge at the time of release. You can see their contents in the
   SHOW DIAL listing. You can change the lists with:

   SET DIAL TONE-COUNTRIES [ cc [ cc [ ... ] ] ]
          Replaces the current TONE-COUNTRIES list with the one given.
          Each cc is a country code; separate them with spaces (not
          commas). Example:

  set dial tone-countries 1 358 44 46 49

          If no country codes are given, the current list, if any, is
          removed, in which case SET DIAL METHOD AUTO is equivalent to SET
          DIAL METHOD DEFAULT.

   SET DIAL PULSE-COUNTRIES [ cc [ cc [ ... ] ] ]
          Replaces the current PULSE-COUNTRIES list with the one give.
          Syntax and operation is like SET DIAL TONE-COUNTRIES.

   If the same country code appears in both lists, Pulse takes precedence.

   The SET DIAL TONE- and PULSE-COUNTRIES commands perform no verification
   whatsoever on the cc's, since almost any syntax might be legal in some
   settings. Furthermore, there is no facility to edit the lists; you can
   only replace the whole list. However, since the only purpose of these
   lists is to establish a basis for picking tone or pulse dialing
   automatically, all you need to override the effect of the list is to
   set a specific dialing method with SET DIAL METHOD TONE or SET DIAL
   METHOD PULSE.

2.1.15. Dial-Modifier Variables

   As of C-Kermit 7.0, dial modifiers are available in the following
   variables:

 \v(dm_lp) Long pause
 \v(dm_sp) Short pause
 \v(dm_pd) Pulse dial
 \v(dm_td) Tone dial
 \v(dm_wa) Wait for answer
 \v(dm_wd) Wait for dialtone
 \v(dm_rc) Return to command mode

   You can use these in your dial strings in place of hardwired modifiers
   like "@", ",", etc, for increased portability of scripts. Example:

  C-Kermit>set modem type usrobotics
  C-Kermit>sho variables dm
   \v(dm_lp) = ,
   \v(dm_sp) = /
   \v(dm_pd) = P
   \v(dm_td) = T
   \v(dm_wa) = @
   \v(dm_wd) = W
   \v(dm_rc) = ;
  C-Kermit>exit

2.1.16. Giving Multiple Numbers to the DIAL Command

   Prior to C-Kermit 7.0, the only way to give a DIAL command a list of
   phone numbers to try until one answers was to create a dialing
   directory that had multiple entries under the same name, and then use
   that entry name in the DIAL command. Now a list of numbers can be given
   to the DIAL command directly in the following format:

  dial {{number1}{number2}{number3}...}

   This is the same list format used by SEND /EXCEPT: and other commands
   that allow a list where normally a single item is given. Restrictions
   on this form of the DIAL command are:

     * The first two braces must be adjacent; spacing is optional
       thereafter.
     * Each number must be an actual number to dial, not a dialing
       directory entry.
     * Dialing directory entries may not contain number lists in this
       format.

   In all other respects, the numbers are treated as if they had been
   fetched from the dialing directory; they can be in literal or portable
   format, etc. Example:

  dial {{7654321} {+1 (212) 5551212} { 1-212-5556789 }}

   The list can be any length at all, within reason.

   This feature is especially handy for use with the K95 Dialer, allowing
   a list of phone numbers to be specified in the Telephone Number box
   without having to set up or reference a separate dialing directory.

   You can also use it to add commonly-dialed sequences as variables in
   your C-Kermit customization file, e.g.:

  define work {{7654321}{7654322}{7654323}}

   and then:

  dial {\m(work)}

   (the variable name must be enclosed in braces).

   Or more simply:

  define work dial {{7654321}{7654322}{7654323}}

   and then:

  work

2.2. Modems

2.2.1. New Modem Types

   Since C-Kermit 6.0:

  atlas-newcom-33600ifxC Atlas/Newcom 33600
  att-keepintouch        AT&T KeepinTouch PCMCIA V.32bis Card Modem
  att-1900-stu-iii       AT&T Secure Data STU-III Model 1900
  att-1910-stu-iii       AT&T Secure Data STU-III Model 1910
  bestdata               Best Data
  cardinal               Cardinal V.34 MVP288X series.
  compaq                 Compaq Data+Fax (e.g. in Presario)
  fujitsu                Fujitsu Fax/Modem Adapter
  generic-high-speed     Any modern error-correcting data-compressing modem
  itu-t-v25ter/v250      ITU-T (CCITT) V.25ter (V.250) standard command set
  megahertz-att-v34      Megahertz AT&T V.34
  megahertz-xjack        Megahertz X-Jack
  motorola-codex         Motorola Codex 326X Series
  motorola-montana       Motorola Montana
  mt5634zpx              Multitech MT5634ZPX
  rockwell-v90           Rockwell V.90 56K
  rolm-244pc             Siemens/Rolm 244PC (AT command set)
  rolm-600-series        Siemens/Rolm 600 Series (AT command set)
  spirit-ii              QuickComm Spirit II
  suprasonic             SupraSonic V288+
  supra-express-v90      Supra Express V.90

   One of the new types, "generic-high-speed" needs a bit of explanation.
   This type was added to easily handle other types that are not
   explicitly covered, without going through the bother of adding a
   complete user-defined modem type. This one works for modern modems that
   use the AT command set, on the assumption that all the default
   ("factory") settings of the modem (a) are appropriate for Kermit, (b)
   include error correction, data compression, and speed buffering; and
   (c) are recallable with the command AT&F.

   If the command to recall your modem's profile is not AT&F, use the SET
   MODEM COMMAND INIT-STRING command to specify the appropriate modem
   command. The default init-string is AT&F\13 (that is, AT, ampersand, F,
   and then carriage return); a survey of about 20 modern modem types
   shows they all support this, but they might mean different things by
   it. For example, the USR Sportster or Courier needs AT&F1 (not AT&F,
   which is equivalent to AT&F0, which recalls an inappropriate profile),
   so for USR modems:

  set modem type generic-high-speed
  set modem command init AT&F1\13

   Of course, USR modems already have their own built-in modem type. But
   if you use this one instead, it will dial faster because it has fewer
   commands to give to the modem; in that sense "&F1" is like a macro that
   bundles numerous commands into a single one. See your modem manual for
   details about factory profiles and commands to recall them.

   WARNING: Do not use the generic-high-speed modem type in operating
   systems like VMS where hardware flow control is not available, at least
   not unless you change the init string from AT&F\13 to something else
   that enables local Xon/Xoff or other appropriate type of flow control.

   Also see Section 2.1.7 for additional hints about making dialing
   go faster.

2.2.2. New Modem Controls

   SET MODEM CAPABILITIES list
          In C-Kermit 7.0, this command automatically turns MODEM
          SPEED-MATCHING OFF if SB (Speed Buffering) is in the list, and
          turns it ON if SB is absent.

   SET MODEM COMMAND PREDIAL-INIT [ text ]
          Commands to be sent to the modem just prior to dialing. Normally
          none.

   SET MODEM SPEAKER { ON, OFF }
          Determines whether modem speaker is on or off while call is
          being placed. ON by default. Note: This command does not provide
          fine-grained control over when the speaker is on or off.
          Normally, ON means while the call is being placed, until the
          point at which carrier is successfully established. If your
          modem has a different speaker option that you want to choose,
          then use the SET MODEM COMMAND SPEAKER ON text command to
          specify this option.

   SET MODEM COMMAND SPEAKER { ON, OFF } [ text ]
          Specify or override the commands to turn your modem's speaker on
          and off.

   SET MODEM VOLUME { LOW, MEDIUM, HIGH }
          When MODEM SPEAKER is on, select volume. Note: In some modems,
          especially internal ones, these commands have no effect; this is
          a limitation of the particular modem, not of Kermit.

   SET MODEM COMMAND VOLUME { LOW, MEDIUM, HIGH } [ text ]
          Specify or override the commands to set your modem's speaker
          volume.

   SET MODEM COMMAND IGNORE-DIALTONE [ text ]
          The command to enable blind dialing ( Section 2.1.6).

   SET MODEM ESCAPE-CHARACTER code
          Has been augmented to allow codes of 0 or less: < 0 means the
          escape mechanism is disabled. = 0 means to use (restore) the
          default value from the modem database. > 0 and < 128 is a
          literal value to be used instead of the default one. > 127 means
          the escape mechanism is disabled. This affects "modem hangup".
          When the escape mechanism is disabled, but SET MODEM
          HANGUP-METHOD is MODEM-COMMAND, it sends the hangup command
          immediately, without the <pause>+++<pause> business first. This
          is useful (for example) when sending lots of numeric pages, a
          process in which never we go online, and so never need to escape
          back. Eliminating the unnecessary pauses and escape sequence
          allows a lot more pages to be sent per unit time.

   Recall that C-Kermit can dial modems to which it is connected via
   TCP/IP (Telnet or Rlogin) as described on page 126 ofUsing
   C-Kermit, 2nd Ed. In this case the MODEM HANGUP-METHOD should be
   MODEM-COMMAND, since RS-232 signals don't work over TCP/IP connections.
   As noted in the manual, such connections are set up by the following
   sequence:

  set host host [ port ]
  set modem type name
  dial number

   But this can cause complications when you use Kermit to switch between
   serial and TCP/IP connections. In the following sequence:

  set host name
  set modem type name
  set port name

   the first two commands obey the rules for dialing out over Telnet.
   However, the SET PORT command requires that Kermit close its current
   (Telnet) connection before it can open the serial port (since Kermit
   can only have one connection open at a time). But since a modem type
   was set after the "set host" command was given, Kermit assumes it is a
   Telnet dialout connection and so sends the modem's hangup sequence is
   sent to the Telnet host. To avoid this, close the network connection
   explicitly before opening the serial one:

  set host name
  close
  set modem type name
  set port name

2.3. TELNET and RLOGIN

   For additional background, please also read theTELNET.TXT file,
   also available on the Web inHTML format.

   Cautions:

     * If making a Telnet connection with C-Kermit takes a very long time,
       like over a minute, whereas the system Telnet program makes the
       same connection immediately, try including the /NOWAIT switch:
  C-Kermit> telnet /nowait hostname

       SeeTELNET.TXT orTELNET.HTM for details. If it also
       takes a very long time to make a Telnet connection with system
       Telnet, then the delay is most likely caused by reverse DNS lookups
       when your host is not properly registered in DNS.
     * When supplying numeric IP addresses to C-Kermit or to any other
       application (regular Telnet, Rlogin, etc), do not include leading
       0's in any fields unless you intend for those fields to be
       interpreted as octal (or hex) numbers. The description of the
       Internet address interpreter (the sockets library inet_addr()
       routine) includes these words:

     All numbers supplied as "parts" in a "." notation may be decimal,
     octal, or hexadecimal, as specified in the C language (that is, a
     leading 0x or 0X implies hexadecimal; otherwise, a leading 0 implies
     octal; otherwise, the number is interpreted as decimal).
       To illustrate, 128.59.39.2 and 128.059.039.002 are not the same
       host! Even though most of the fields contain non-octal digits.
       Using system Telnet (not Kermit):
  $ telnet 128.059.039.002
  Trying 128.49.33.2 ...

       Of course the same thing happens with Kermit because it uses (as it
       must) the same system service for resolving network addresses that
       Telnet, FTP, and all other TCP/IP applications use.
     * The RLOGIN section on page 123 does not make it clear that you can
       use the SET TELNET TERMINAL-TYPE command to govern the terminal
       type that is reported by C-Kermit to the RLOGIN server.
     * Note that the SET TCP commands described on pages 122-123 might be
       absent; some platforms that support TCP/IP do not support these
       particular controls.

   New commands:

   TELOPT { AO, AYT, BREAK, CANCEL, EC, EL, EOF, EOR, GA, IP, DMARK,
          DO, DONT, NOP, SB, SE, SUSP, WILL, WONT }
          This command was available previously, but supported only DO,
          DONT, WILL, and WONT. Now it lets you send all the Telnet
          protocol commands. Note that certain commands do not require a
          response, and therefore can be used as nondestructive "probes"
          to see if the Telnet session is still open; e.g.:

  set host xyzcorp.com
  ...
  telopt nop
  if fail stop 1 Connection lost

   SET TCP ADDRESS [ ip-address ]
          Specifies the IP address of the computer that C-Kermit is
          running on. Normally this is not necessary. The exception would
          be if your machine has multiple network adapters (physical or
          virtual) with a different address for each adapter AND you want
          C-Kermit to use a specific address when making outgoing
          connections or accepting incoming connections.

   SET TCP DNS-SERVICE-RECORDS { ON, OFF }
          Tells C-Kermit whether to try to use DNS SRV records to
          determine the host and port number upon which to find an
          advertised service. For example, if a host wants regular Telnet
          connections redirected to some port other than 23, this feature
          allows C-Kermit to ask the host which port it should use. Since
          not all domain servers are set up to answer such requests, this
          feature is OFF by default.

   SET TCP REVERSE-DNS-LOOKUP { ON, OFF, AUTO }
          Tells Kermit whether to perform a reverse DNS lookup on TCP/IP
          connections. This allows Kermit to determine the actual hostname
          of the host it is connected to, which is useful for connections
          to host pools, and is required for Kerberos connections to host
          pools and for incoming connections. If the other host does not
          have a DNS entry, the reverse lookup could take a long time
          (minutes) to fail, but the connection will still be made. Turn
          this option OFF for speedier connections if you do not need to
          know exactly which host you are connected to and you are not
          using Kerberos. AUTO, the default, means the lookup is done on
          hostnames, but not on numeric IP addresses.

   SET TELNET WAIT-FOR-NEGOTIATIONS { ON, OFF }
          Each Telnet option must be fully negotiated either On or Off
          before the session can continue. This is especially true with
          options that require sub-negotiations such as Authentication,
          Encryption, and Kermit; for proper support of these options
          Kermit must wait for the negotiations to complete. Of course,
          Kermit has no way of knowing whether a reply is delayed or not
          coming at all, and so will wait a minute or more for required
          replies before continuing the session. If you know that Kermit's
          Telnet partner will not be sending the required replies, you can
          set this option of OFF to avoid the long timeouts. Or you can
          instruct Kermit to REFUSE specific options with the SET TELOPT
          command.

   SET TELOPT [ { /CLIENT, /SERVER } ] option
          { ACCEPTED, REFUSED, REQUESTED, REQUIRED }
          [ { ACCEPTED, REFUSED, REQUESTED, REQUIRED } ]
          SET TELOPT lets you specify policy requirements for Kermit's
          handling of Telnet option negotiations. Setting an option is
          REQUIRED causes Kermit to offer the option to the peer and
          disconnect if the option is refused. REQUESTED causes Kermit to
          offer an option to the peer. ACCEPTED results in no offer but
          Kermit will attempt to negotiate the option if it is requested.
          REFUSED instructs Kermit to refuse the option if it is requested
          by the peer.

          Some options are negotiated in two directions and accept
          separate policies for each direction; the first keyword applies
          to Kermit itself, the second applies to Kermit's Telnet partner;
          if the second keyword is omitted, an appropriate
          (option-specific) default is applied. You can also include a
          /CLIENT or /SERVER switch to indicate whether the given policies
          apply when Kermit is the Telnet client or the Telnet server; if
          no switch is given, the command applies to the client.

          Note that some of Kermit's Telnet partners fail to refuse
          options that they do not recognize and instead do not respond at
          all. In this case it is possible to use SET TELOPT to instruct
          Kermit to REFUSE the option before connecting to the problem
          host, thus skipping the problematic negotiation.

          Use SHOW TELOPT to view current Telnet Option negotiation
          settings. SHOW TELNET displays current Telnet settings.

2.3.0. Bug Fixes

   If "set host nonexistent-host" was given (and it properly failed),
   followed by certain commands like SEND, the original line and modem
   type were not restored and C-Kermit thought that it still had a network
   hostname; fixed in 7.0.

2.3.1. Telnet Binary Mode Bug Adjustments

   SET TELNET BUG BINARY-ME-MEANS-U-TOO { ON, OFF } was added to edit 192
   after the book was printed. Also SET TELNET BUG BINARY-U-MEANS-ME-TOO.
   The default for both is OFF. ON should be used when communicating with
   a Telnet partner (client or server) that mistakenly believes that
   telling C-Kermit to enter Telnet binary mode also means that it, too,
   is in binary mode, contrary to the Telnet specification, which says
   that binary mode must be negotiated in each direction separately.

2.3.2. VMS UCX Telnet Port Bug Adjustment

   A new command, SET TCP UCX-PORT-BUG, was added for VMS versions with
   UCX (DEC TCP/IP), applying only to early versions of UCX, like 2.2 or
   earlier. If you try to use VMS C-Kermit to make a Telnet connection
   using a port name (like "telnet", which is used by default), the
   underlying UCX getservbyname() function might return the service number
   with its bytes swapped and the connection will fail. If "telnet
   hostname 23" works, then your version of UCX has this bug and you can
   put "set tcp ucx-port-bug on" in your CKERMIT.INI file to get around
   it.

2.3.3. Telnet New Environment Option

   The TELNET NEW-ENVIRONMENT option (RFC 1572) is supported as 7.0.
   This option allows the C-Kermit Telnet client to send certain
   well-known variables to the Telnet server, including USER, PRINTER,
   DISPLAY, and several others. This feature is enabled by default in
   Windows and OS/2, disabled by default elsewhere. The command to enable
   and disable it is:

  SET TELNET ENVIRONMENT { ON, OFF }

   When ON, and you Telnet to another computer, you might (or might not)
   notice that the "login:" or "Username:" prompt does not appear --
   that's because your username was sent ahead, in which case the remote
   system might prompt you only for your password (similar to Rlogin). Use
   "set telnet environment off" to defeat this feature, particularly in
   scripts where the dialog must be predictable. You can also use this
   command to specify or override specific well-known environment variable
   values:

 SET TELNET ENVIRONMENT { ACCT,DISPLAY,JOB,PRINTER,SYSTEMTYPE,USER } [ text ]

2.3.4. Telnet Location Option

   The TELNET LOCATION option (RFC 779) is supported in 7.0. This
   option allows the C-Kermit Telnet client to send a location string to
   the server if the server indicates its willingness to accept one. If an
   environment variable named LOCATION exists at the time C-Kermit starts,
   its value is used as the location string. If you want to change it,
   use:

  SET TELNET LOCATION text

   If you omit the text from this command, the Telnet location feature is
   disabled.

   SET TELNET ENVIRONMENT DISPLAY is used to set the DISPLAY variable that
   is sent to the host, as well as the XDISPLAY location.

2.3.5. Connecting to Raw TCP Sockets

   The SET HOST and TELNET commands now accept an optional switch,
   /RAW-SOCKET, at the end, only if you first give a host and a port.
   Example:

  set host xyzcorp.com 23 /raw-socket
  set host 128.49.39.2:2000 /raw-socket
  telnet xyzcorp.com 3000 /raw

   Without this switch, C-Kermit behaves as a Telnet client when (a) the
   port is 23 or 1649, or (b) the port is not 513 and the server sent what
   appeared to be Telnet negotiations -- that is, messages starting with
   0xFF (IAC). With this switch, Kermit should treat all incoming bytes as
   raw data, and will not engage in any Telnet negotiations or NVT CRLF
   manipulations. This allows transparent operation through (e.g.) raw TCP
   ports on Cisco terminal servers, through the 'modemd' modem server,
   etc.

2.3.6. Incoming TCP Connections

   Accomplished via SET HOST * port, were introduced in C-Kermit 6.0, but
   for UNIX only. In Version 7.0, they are also available for VMS.

2.4. The EIGHTBIT Command

   EIGHTBIT is simply a shorthand for: SET PARITY NONE, SET TERMINAL
   BYTESIZE 8, SET COMMAND BYTESIZE 8; that is, a way to set up an 8-bit
   clean connection in a single command.

2.5. The Services Directory

   Chapter 7 ofUsing C-Kermit does not mention the ULOGIN macro,
   which is used by our sample services directory, CKERMIT.KND. Unlike
   UNIXLOGIN, VMSLOGIN, etc, this one is for use with systems that require
   a user ID but no password. Therefore it doesn't prompt for a password
   or wait for a password prompt from the remote service.

   In version 7.0, the CALL macro was changed to not execute a SET MODEM
   TYPE command if the given modem type was the same as the current one;
   otherwise the new SET MODEM TYPE command would overwrite any
   customizations that the user had made to the modem settings. Ditto for
   SET LINE / SET PORT and SET SPEED.

2.6. Closing Connections

   Until version 7.0, there was never an obvious and general way to close
   a connection. If a serial connection was open, it could be closed by
   "set line" or "set port" (giving no device name); if a network
   connection was open, it could be closed by "set host" (no host name).

   In version 7.0, a new command closes the connection in an obvious and
   straightforward way, no matter what the connection type:

  CLOSE [ CONNECTION ]

   The CLOSE command was already present, and required an operand such as
   DEBUG-LOG, WRITE-FILE, etc, and so could never be given by itself. The
   new CONNECTION operand is now the default operand for CLOSE, so CLOSE
   by itself closes the connection, if one is open, just as you would
   expect, especially if you are a Telnet or Ftp user.

   Also see the description of the new SET CLOSE-ON-DISCONNECT command in
  Section 2.10.

2.7. Using C-Kermit with External Communication Programs

   C-Kermit 7.0 includes a new ability to create and conduct sessions
   through other communications programs. Two methods are available:

    1. Pty (pseudoterminal): The external program is run on a
       "pseudoterminal", which is controlled by Kermit. This method works
       with practically any external program, but it is not portable. At
       this writing, it works only on some (not all) UNIX versions, and
       not on any non-UNIX platforms.
    2. Pipe: The external program's standard input and output are
       redirected through a "pipe" controlled by Kermit. This method is
       relatively portable -- it should work across all UNIX versions, and
       it also works in Windows and OS/2 -- but it is effective only when
       the external program actually uses standard i/o (and many don't).

   The two methods are started differently but are used the same way
   thereafter.

   The purpose of this feature is to let you use C-Kermit services like
   file transfer, character-set translation, scripting, automatic dialing,
   etc, on connections that Kermit can't otherwise make itself.

   This feature is the opposite of the REDIRECT feature, in which C-Kermit
   makes the connection, and redirects an external (local) command or
   program over this connection. In a pty or pipe connection, C-Kermit
   runs and controls a local command or program, which makes the
   connection. (The same method can be used to simply to control a local
   program without making a connection; see Section 2.8.)

   To find out if your version of Kermit includes PTY support, type "show
   features" and look for NETPTY in the alphabetical list of options. For
   pipes, look for NETCMD.

   The commands are:

   SET NETWORK TYPE PTY or SET NETWORK TYPE PIPE
   SET HOST command
          where command is any interactive command. If the command does
          not use standard i/o, you must use SET NETWORK TYPE PTY.

   Notes:

     * COMMAND is an invisible synonym for PIPE.
     * The command and its arguments are case-sensitive in UNIX.

   The SET NETWORK TYPE, SET HOST sequence sets the given network type for
   all subsequent SET HOST commands until another SET NETWORK TYPE command
   is given to change it.

   You can also use the new /NETWORK-TYPE:PTY or /NETWORK-TYPE:PIPE (or
   simply /PIPE or /PTY) switches on the SET HOST command itself:

  SET HOST /NETWORK-TYPE:PIPE command  ; These two are the same
  SET HOST /PIPE command

  SET HOST /NETWORK-TYPE:PTY command   ; Ditto
  SET HOST /PTY command

   These are like SET NETWORK TYPE followed by SET HOST, except they apply
   only to the connection being made and do not change the global network
   type setting (see Section 1.5 about the difference between
   switches and SET commands).

   Include any command-line options with the command that might be needed,
   as in this example where C-Kermit uses another copy of itself as the
   communications program:

  SET HOST /PIPE /CONNECT kermit -YQJ xyzcorp.com

     IMPORTANT: In Unix, wildcards and redirectors are interpreted by the
     shell. If you want to run a program with (say) SET HOST /PTY with
     its i/o redirected or with wildcard file arguments, you will need to
     invoke the shell too. Example:

SET HOST /PTY {sh -c "crypt < foo.x"}
SET HOST /PTY {sh -c "grep somestring *.txt"}

   As usual, if you include the /CONNECT switch, SET HOST enters CONNECT
   mode immediately upon successful execution of the given command.
   Therefore new commands are available as a shorthand for SET HOST
   /CONNECT /PTY and /PIPE:

   PTY [ command ]
   PIPE [ command ]
          The PTY and PIPE commands work like the TELNET and RLOGIN
          commands: they set up the connection (in this case, using the
          given command) and then enter CONNECT mode automatically (if the
          PIPE or PTY command is given without a command, it continues the
          current session if one is active; otherwise it gives an error
          message).

   The PIPE command is named after the mechanism by which C-Kermit
   communicates with the command: UNIX pipes. C-Kermit's i/o is "piped"
   through the given command. Here is a typical example:

  PIPE rlogin -8 xyzcorp.com

   This is equivalent to:

  SET HOST /PIPE rlogin -8 xyzcorp.com
  CONNECT

   and to:

  SET HOST /PIPE /CONNECT rlogin -8 xyzcorp.com

   IMPORTANT:
          If you are writing a script, do not use the PIPE, PTY, TELNET,
          or RLOGIN command unless you really want C-Kermit to enter
          CONNECT mode at that point. Normally SET HOST is used in scripts
          to allow the login and other dialogs to be controlled by the
          script itself, rather than by an actively participating human at
          the keyboard.

   Throughput of pty and pipe connections is limited by the performance of
   the chosen command or program and by the interprocess communication
   (IPC) method used and/or buffering capacity of the pipe or pty, which
   in turn depends on the underlying operating system.

   In one trial (on SunOS 4.1.3), we observed file transfer rates over an
   rlogin connection proceeding at 200Kcps for downloads, but only 10Kcps
   for uploads on the same connection with the same settings (similar
   disparities were noted in HP-UX). Examination of the logs revealed that
   a write to the pipe could take as long as 5 seconds, whereas reads were
   practically instantaneous. On the other hand, using Telnet as the
   external program rather than rlogin, downloads and uploads were better
   matched at about 177K each.

   Most external communication programs, like C-Kermit itself, have escape
   characters or sequences. Normally these begin with (or consist entirely
   of) a control character. You must be sure that this control character
   is not "unprefixed" when uploading files, otherwise the external
   program will "escape back" to its prompt, or close the connection, or
   take some other unwanted action. When in CONNECT mode, observe the
   program's normal interaction rules. Of course C-Kermit's own escape
   character (normally Ctrl-\) is active too, unless you have taken some
   action to disable it.

   On PTY connections, the underlying PTY driver is not guaranteed to be
   transparent to control characters -- for example, it might expand tabs,
   translate carriage returns, generate signals if it sees an interrupt
   character, and so on. Similar things might happen on a PIPE connection.
   For this reason, if you plan to transfer files over a PTY or PIPE
   connection, tell the file sender to:

   SET PREFIXING ALL
          This causes all control characters to be prefixed and
          transmitted as printable ASCII characters.

   If the external connection program is not 8-bit clean, you should also:

   SET PARITY SPACE
          This causes 8-bit data to be encoded in 7 bits using single
          and/or locking shifts.

   And if it does not make a reliable connection (such as those made by
   Telnet, Rlogin, Ssh, etc), you should:

   SET STREAMING OFF
          This forces C-Kermit to treat the connection as unreliable and
          to engage in its normal ACK/NAK protocol for error detection and
          correction, rather than "streaming" its packets, as it normally
          does on a network connection ( Section 4.20).

   In some cases, buffer sizes might be restricted, so you might also need
   to reduce the Kermit packet length to fit; this is a trial-and-error
   affair. For example, if transfers always fail with 4000-byte packets,
   try 2000. If that fails too, try 1000, and so on. The commands are:

   SET RECEIVE PACKET-LENGTH number
          This tells the file receiver to tell the file sender the longest
          packet length it can accept.

   SET SEND PACKET-LENGTH number
          This tells the file sender not to send packets longer than the
          given length, even if the receiver says longer ones are OK. Of
          course, if the receiver's length is shorter, the shorter length
          is used.

   If none of this seems to help, try falling back to the bare minimum,
   lowest-common-denominator protocol settings:

   ROBUST
          No sliding windows, no streaming, no control-character
          unprefixing, packet length 90.

   And then work your way back up by trial and error to get greater
   throughput.

   Note that when starting a PIPE connection, and the connection program
   (such as telnet or rlogin) prints some greeting or information messages
   before starting the connection, these are quite likely to be printed
   with a stairstep effect (linefeed without carriage return). This is
   because the program is not connected with the UNIX terminal driver;
   there's not much Kermit can do about it. Once the connection is made,
   everything should go back to normal. This shouldn't happen on a PTY
   connection because a PTY is, indeed, a terminal.

   On a similar note, some connection programs (like Solaris 2.5 rlogin)
   might print lots of error messages like "ioctl TIOCGETP: invalid
   argument" when used through a pipe. They are annoying but usually
   harmless. If you want to avoid these messages, and your shell allows
   redirection of stderr, you can redirect stderr in your pipe command, as
   in this example where the user's shell is bash:

  PIPE rlogin xyzcorp.com 2> /dev/null

   Or use PTY rather than PIPE, since PTY is available on Solaris.

2.7.0. C-Kermit over tn3270 and tn5250

   Now you can make a connection from C-Kermit "directly" to an IBM
   mainframe and transfer files with it, assuming it has Kermit-370
   installed. Because tn3270 is neither 8-bit clean nor transparent to
   control characters, you must give these commands:

  SET PREFIXING ALL   ; Prefix all control characters
  SET PARITY SPACE    ; Telnet connections are usually not 8-bit clean

   and then:

  SET HOST /PTY /CONNECT tn3270 abccorp.com

   or simply:

  pty tn3270 abccorp.com

   SET HOST /PIPE does not work in this case, at least not for file
   transfer. File transfer does work, however, with SET HOST /PTY,
   provided you use the default packet length of 90 bytes; anything longer
   seems to kill the session.

   You can also make connections to IBM AS/400 computers if you have a
   tn5250 program installed:

  pty tn5250 hostname

   In this case, however, file transfer is probably not in the cards since
   nobody has ever succeeded in writing a Kermit program for the AS/400.
   Hint:

  define tn3270 {
      check pty
      if fail end 1 Sorry - no PTY support...
      pty tn3270 \%*
  }

   Similarly for tn5250. Note that CHECK PTY and CHECK PIPE can be used in
   macros and scripts to test whether PTY or PIPE support is available.

2.7.1. C-Kermit over Telnet

   Although C-Kermit includes its own Telnet implementation, you might
   need to use an external Telnet program to make certain connections;
   perhaps because it has access or security features not available in
   C-Kermit itself. As noted above, the only precautions necessary are
   usually:

  SET PREFIXING ALL   ; Prefix all control characters
  SET PARITY SPACE    ; Telnet connections might not be 8-bit clean

   and then:

  SET HOST /PTY (or /PIPE) /CONNECT telnet abccorp.com

   or, equivalently:

  PTY (or PIPE) telnet abccorp.com

2.7.2. C-Kermit over Rlogin

   C-Kermit includes its own Rlogin client, but this can normally be used
   only if you are root, since the rlogin TCP port is privileged. But ptys
   and pipes let you make rlogin connections with C-Kermit through your
   computer's external rlogin program, which is normally installed as a
   privileged program:

  SET PREFIXING ALL

   and then:

  SET HOST /PTY (or /PIPE) /CONNECT rlogin -8 abccorp.com

   or, equivalently:

  PTY (or PIPE) rlogin -8 abccorp.com

   The "-8" option to rlogin enables transmission of 8-bit data. If this
   is not available, then include SET PARITY SPACE if you intend to
   transfer files.

   Note that the normal escape sequence for rlogin is Carriage Return
   followed by Tilde (~), but only when the tilde is followed by certain
   other characters; the exact behavior depends on your rlogin client, so
   read its documentation.

2.7.3. C-Kermit over Serial Communication Programs

   Ptys and pipes also let you use programs that make serial connections,
   such as cu or tip. For example, C-Kermit can be used through cu to make
   connections that otherwise might not be allowed, e.g. because C-Kermit
   is not installed with the required write permissions to the dialout
   device and the UUCP lockfile directory.

   Suppose your UUCP Devices file contains an entry for a serial device
   tty04 to be used for direct connections, but this device is protected
   against you (and Kermit when you run it). In this case you can:

  SET CONTROL PREFIX ALL
  PTY (or PIPE) cu -l tty04

   (Similarly for dialout devices, except then you also need to include
   the phone number in the "cu" command.)

   As with other communication programs, watch out for cu's escape
   sequence, which is the same as the rlogin program's: Carriage Return
   followed by Tilde (followed by another character to specify an action,
   like "." for closing the connection and exiting from cu).

2.7.4. C-Kermit over Secure Network Clients

     DISCLAIMER: There are laws in the USA and other countries regarding
     use, import, and/or export of encryption and/or decryption or other
     forms of security software, algorithms, technology, and intellectual
     property. The Kermit Project attempts to follow all known statutes,
     and neither intends nor suggests that Kermit software can or should
     be used in any way, in any location, that circumvents any
     regulations, laws, treaties, covenants, or other legitimate canons
     or instruments of law, international relations, trade, ethics, or
     propriety.

   For secure connections or connections through firewalls, C-Kermit 7.0
   can be a Kerberos, SRP, and/or SOCKS client when built with the
   appropriate options and libraries. But other application-level security
   acronyms and methods -- SSH, SSL, SRP, TLS -- pop up at an alarming
   rate and are (a) impossible to keep up with, (b) usually mutually
   incompatible, and (c) have restrictions on export or redistribution and
   so cannot be included in C-Kermit itself.

   However, if you have a secure text-based Telnet (or other) client that
   employs one of these security methods, you can use C-Kermit "through"
   it via a pty or pipe.

2.7.4.1. SSH

   C-Kermit does not and can not incorporate SSH due to licensing, patent,
   and USA export law restrictions.

   The UNIX SSH client does not use standard input/output, and therefore
   can be used only by Kermit's PTY interface, if one is present. The
   cautions about file transfer, etc, are the same as for Rlogin. Example:

  SET PREFIXING ALL
  PTY ssh XYZCORP.COM

   Or, for a scripted session:

  SET PREFIXING ALL
  SET HOST /PTY ssh XYZCORP.COM

   Hint:

  define ssh {
      check pty
      if fail end 1 Sorry - no PTY support...
      pty ssh \%*
  }

2.7.4.2. SSL

   Secure Sockets Layer (SSL) is another TCP/IP security overlay, this one
   designed by and for Netscape. An SSL Telnet client is available for
   UNIX from the University of Queensland. More info at:

 http://www.psy.uq.oz.au/~ftp/Crypto/

   Interoperability with C-Kermit is unknown. C-Kermit also includes its
   own built-in SSL/TLS support, but it is not exportable.

2.7.4.3. SRP

   SRP(TM) is Stanford University's Secure Remote Password protocol. An
   SRP Telnet client is available from Stanford:

 http://srp.stanford.edu/srp/

   Stanford's SRP Telnet client for UNIX has been tested on SunOS and
   works fine with C-Kermit, as described in Section 2.7.1, e.g.

  SET PREFIX ALL
  PTY (or PIPE) srp-telnet xenon.stanford.edu

   C-Kermit itself can be built as an SRP Telnet client on systems that
   have libsrp.a installed; the C-Kermit support code, however, may not be
   exported outside the USA or Canada.

2.7.4.4. SOCKS

   C-Kermit can be built as a SOCKS-aware client on systems that have a
   SOCKS library. See section 8.1.1 of theckccfg.txt file.

   C-Kermit 7.0 can also be run over SOCKSified Telnet or rlogin clients
   with SET NETWORK TYPE COMMAND. Suppose the Telnet program on your
   system is SOCKS enabled but C-Kermit is not. Make Kermit connections
   like this:

  SET PREFIX ALL
  PTY (or PIPE) telnet zzz.com

2.7.4.5. Kerberos

   UNIX C-Kermit can be built with MIT Kerberos IV or V authentication and
   encryption. Instructions are available in aseparate document.
   Additional modules are required that can not be exported from the USA
   to any country except Canada, by US law.

   If you have Kerberos installed but you don't have a Kerberized version
   of C-Kermit, you can use ktelnet as C-Kermit's external communications
   program to make secure connections without giving up C-Kermit's
   services:

  SET PREFIX ALL
  PTY (or PIPE) ktelnet cia.gov

2.8. Scripting Local Programs

   If your version of Kermit has PTY support built in, then any text-based
   program can be invoked with SET HOST /PTY or equivalent command and
   controlled using the normal sequence of OUTPUT, INPUT, IF SUCCESS
   commands (this is the same service that is provided by the 'expect'
   program, but controlled by the Kermit script language rather than Tcl).

   When PTY service is not available, then any program that uses standard
   input and output can be invoked with SET HOST /PIPE.

   Here's an example in which we start an external Kermit program, wait
   for its prompt, give it a VERSION command, and then extract the numeric
   version number from its response:

  set host /pty kermit -Y
  if fail stop 1 {Can't start external command}
  input 10 C-Kermit>
  if fail stop 1 {No C-Kermit> prompt}
  output version\13
  input 10 {Numeric: }
  if fail stop 1 {No match for "Numeric:"}
  clear input
  input 10 \10
  echo VERSION = "\fsubstr(\v(input),1,6)"
  output exit\13

   This technique could be used to control any other interactive program,
   even those that do screen formatting (like Emacs or Vi), if you can
   figure out the sequence of events. If your Kermit program doesn't have
   PTY support, then the commands are restricted to those using standard
   i/o, including certain shells, interactive text-mode "hardcopy" editors
   like ex, and so on.

   If you are using the PTY interface, you should be aware that it runs
   the given program or command directly on the pty, without any
   intervening shell to interpret metacharacters, redirectors, etc. If you
   need this sort of thing, include the appropriate shell invocation as
   part of your command; for example:

  pty echo *

   just echoes "*"; whereas:

  pty ksh -c "echo *"

   echoes all the filenames that ksh finds matching "*".

   Similarly for redirection:

  set host /pty ksh -c "cat > foo"  ; Note: use shell quoting rules here
  set transmit eof \4
  transmit bar

   And for that matter, for built-in shell commands:

  set host /pty ksh -c "for i in *; do echo $i; done"

   The PIPE interface, on the other hand, invokes the shell automatically,
   so:

  pipe echo *

   prints filenames, not "*".

2.9. X.25 Networking

   X.25 networking is documented inUsing C-Kermit, 2nd Edition. When
   the book was published, X.25 was available only in SunOS, Solaris, and
   Stratus VOS. Unlike TCP/IP, X.25 APIs are not standardized; each
   vendor's X.25 libraries and services (if they have them at all) are
   unique.

   This section describes new additions.

2.9.1. IBM AIXLink/X.25 Network Provider Interface for AIX

   Support for X.25 was added via IBM's Network Provider Interface (NPI),
   AIXLink/X.25 1.1, to the AIX 4.x version of C-Kermit 7.0.
   Unfortunately, AIXLink/X.25 is a rather bare-bones facility, lacking in
   particular any form of PAD support (X.3, X.28, X.29). Thus, the AIX
   version of C-Kermit, when built to include X.25 networking, has neither
   a PAD command, nor a SET PAD command. The same is true for the
   underlying AIX system: no PAD support. Thus it is not possible to have
   an interactive shell session over an X.25 connection into an AIX system
   (as far as we know), even from X.25-capable Kermit versions (such as
   Solaris or VOS) that do include PAD support.

   Thus the X.25 capabilities in AIX C-Kermit are limited to peer-to-peer
   connections, e.g. from a C-Kermit client to a C-Kermit server. Unlike
   the Solaris, SunOS, and VOS versions, the AIX version can accept
   incoming X.25 connections:

  set network type x.25
  if fail stop 1 Sorry - no X.25 support
  ; Put any desired DISABLE or ENABLE or SET commands here.
  set host /server *
  if fail stop 1 X.25 "set host *" failed

   And then access it from the client as follows:

  set network type x.25
  if fail stop 1 Sorry - no X.25 support
  set host xxxxxxx ; Specify the X.25/X.121 address
  if fail stop 1 Can't open connection

   And at this point the client can use the full range of client commands:
   SEND, GET, REMOTE xxx, FINISH, BYE.

   The AIX version also adds two new variables:

   \v(x25local_nua)
          The local X.25 address.

   \v(x25remote_nua)
          The X.25 address of the host on the other end of the connection.

   C-Kermit's AIX X.25 client has not been tested against anything other
   than a C-Kermit X.25 server on AIX. It is not known if it will
   interoperate with C-Kermit servers on Solaris, SunOS, or VOS.

   To make an X.25 connection from AIX C-Kermit, you must:

  set x25 call-user-data xxxx

   where xxxx can be any even-length string of hexadecimal digits, e.g.
   123ABC.

2.9.2. HP-UX X.25

   Although C-Kermit presently does not include built-in support for HP-UX
   X.25, it can still be used to make X.25 connections as follows: start
   Kermit and tell it to:

  set prefixing all
  set parity space
  pty padem address

   This should work in HP-UX 9.00 and later (see Section 2.7). If you
   have an earlier HP-UX version, or the PTY interface doesn't work or
   isn't available, try:

  set prefixing all
  set parity space
  pipe padem address

   Failing that, use Kermit to telnet to localhost and then after logging
   back in, start padem as you would normally do to connect over X.25.

2.10. Additional Serial Port Controls

   C-Kermit 7.0 adds the following commands for greater control over
   serial ports. These commands are available only in C-Kermit versions
   whose underlying operating systems provide the corresponding services
   (such as POSIX and UNIX System V), and even then their successful
   operation depends on the capabilities of the specific device and
   driver.

   SET DISCONNECT { ON, OFF }
          On a SET LINE or SET PORT connection with SET CARRIER ON or
          AUTO, if the carrier signal drops during the connection,
          indicating that the connection has been lost, and C-Kermit
          notices it, this setting governs what happens next. With SET
          DISCONNECT OFF, which is consistent with previous behavior, and
          therefore the default, C-Kermit continues to keep the device
          open and allocated. With SET DISCONNECT ON, C-Kermit
          automatically closes and releases the device when it senses a
          carrier on-to-off transition, thus allowing others to use it.
          However, it remains the default device for i/o (DIAL, REDIAL,
          INPUT, SEND, CONNECT, etc), so if a subsequent i/o command is
          given, the device is reopened if it is still available. When it
          has been automatically closed in this manner, SHOW
          COMMUNICATIONS puts "(closed)" after its name, and in UNIX, the
          lockfile disappears -- both from SHOW COMM and from the lockfile
          directory itself. Synonym: SET CLOSE-ON-DISCONNECT.

   SET EXIT ON-DISCONNECT { ON, OFF }
          Like DISCONNECT, but makes the program exit if a connection
          drops.

   Note that SET CLOSE-ON-DISCONNECT and SET EXIT ON-DISCONNECT apply only
   to connections that drop; they do not apply to connections that can't
   be made in the first place. For example, they have no effect when a SET
   LINE, SET HOST, TELNET, or DIAL command fails.

   HANGUP
          If [CLOSE-ON-]DISCONNECT is ON, and the HANGUP command is given
          on a serial device, and the carrier signal is no longer present
          after the HANGUP command, the device is closed and released.

   SET PARITY HARDWARE { EVEN, ODD }
          Unlike SET PARITY { EVEN, ODD, MARK, SPACE }, which selects 7
          data bits plus the indicated kind of parity (to be done in
          software by Kermit itself), SET PARITY HARDWARE selects 8 data
          bits plus even or odd parity, to be done by the underlying
          hardware, operating system, or device driver. This command is
          effective only with a SET LINE or SET PORT device. That is, it
          has no effect in remote mode, nor on network connections. There
          is presently no method for selecting 8 data bits plus mark or
          space parity. If hardware parity is in effect, the variable
          \v(hwparity) is set to "even" or "odd". Note: some platforms
          might also support settings of SPACE, MARK, or NONE.

   SET STOP-BITS { 1, 2 }
          This tells the number of 1-bits to insert after an outbound
          character's data and parity bits, to separate it from the next
          character. Normally 1. Choosing 2 stop bits should do no harm,
          but will slow down serial transmission by approximately 10
          percent. Historically, 2 stop bits were used with Teletypes (at
          110 bps or below) for print-head recovery time. There is
          presently no method for choosing any number of stop bits besides
          1 and 2.

   SET SERIAL [ dps ]
          dps stands for Data-bits, Parity, Stop-bits. This is the
          notation familiar to many people for serial port configuration:
          7E1, 8N1, 7O2, etc. The data bits number also becomes the
          TERMINAL BYTESIZE setting. The second character is E for Even, O
          for Odd, M for Mark, S for Space, or N for None. The list of
          available options depends on the capabilities of the specific
          platform. If dps is omitted, 8N1 is used. Type "set serial ?"
          for a list of available choices. Examples:

        SET SERIAL 7E1
                Equivalent to SET PARITY EVEN, SET STOP-BITS 1, SET TERM
                BYTE 7.

        SET SERIAL 8N1
                Equivalent to SET PARITY NONE, SET STOP-BITS 1, SET TERM
                BYTE 8.

        SET SERIAL 7E2
                Equivalent to SET PARITY EVEN and SET STOP-BITS 2, SET
                TERM BYTE 7.

        SET SERIAL 8E2
                Same as SET PARITY HARDWARE EVEN, SET STOP-BITS 2, SET
                TERM BYTE 8.

        SET SERIAL
                Same as SET PARITY NONE and SET STOP-BITS 1, SET TERM BYTE
                8.

   Notes:

     * The SET SERIAL xx2 options are available only in Kermit versions
       where the SET PARITY HARDWARE command is also available. (SHOW
       FEATURES includes "HWPARITY" in its options list.)
     * The SET SERIAL 7xx and 8N1 options affect the software parity
       setting, even for network connections.
     * As noted in the manual, selecting 8 data bits does not give you
       8-bit terminal sessions in CONNECT mode unless you also SET
       TERMINAL BYTESIZE 8. The default terminal bytesize remains 7, to
       protect against the situation where the remote host is generating
       parity but you don't know about it. If the terminal bytesize was 8
       by default and you CONNECTed to such a host, you would see only
       garbage on your screen.
     * If you do not give a SET STOP-BITS or SET SET SERIAL command,
       C-Kermit does not attempt to set the device's stop bits; instead,
       it uses whatever setting the device uses when not given explicit
       instructions about stop bits.

   SHOW COMMUNICATIONS displays the current settings. Stop bits and
   hardware parity are shown only for SET PORT / SET LINE (serial)
   devices, since they do not apply to network connections or to remote
   mode. STOP-BITS is shown as "(default)" if you have not given an
   explicit SET STOP-BITS or SET SERIAL command.

   The \v(serial) variable shows the SET SERIAL setting (8N1, 7E1, etc).

2.11. Getting Access to the Dialout Device

     This section is for UNIX only; note the special words about QNX at
     the end. Also see Section 2.0 for SET LINE switches,
     particularly the /SHARE switch for VMS only.

   C-Kermit does its best to obey the UUCP lockfile conventions of each
   platform (machine, operating system, OS version) where it runs, if that
   platform uses UUCP.

   But simply obeying the conventions is often not good enough, due to the
   increasing likelihood that a particular serial device might have more
   than one name (e.g. /dev/tty00 and /dev/term/00 are the same device in
   Unixware 7; /dev/cua and /dev/cufa are the same device in NeXTSTEP),
   plus the increasingly widespread use of symlinks for device names, such
   as /dev/modem.

   C-Kermit 7.0 goes to greater lengths than previous versions to
   successfully interlock with other communications program (and other
   instances of Kermit itself); for example, by:

     * Creation of dual lockfiles whenever a symlink is used; one for the
       link name and one for the real name.
     * Creation of dual lockfiles in HP-UX according to HP rules.
     * Creation of dual uppercase/lowercase lockfile names in SCO
       UNIX/ODT/OSR5.
     * The use of ttylock() in versions of AIX where it works.
     * The use, wherever possible, of lockfile names based on
       inode/major/minor device number rather than device name.

   See theckuins.txt andckubwr.txt files for details.

   QNX is almost unique among UNIX varieties in having no UUCP programs
   nor UUCP-oriented dialout-device locking conventions. QNX does,
   however, allow a program to get the device open count. This can not be
   a reliable form of locking unless all applications do it (and they
   don't), so by default, Kermit uses this information only for printing a
   warning message such as:

  C-Kermit>set line /dev/ser1
  WARNING - "/dev/ser1" looks busy...

   However, if you want to use it as a lock, you can do so with:

  SET QNX-PORT-LOCK { ON, OFF }

   QNX-PORT-LOCK is OFF by default; if you set in ON, C-Kermit fails to
   open any dialout device when its open count indicates that another
   process has it open. SHOW COMM (in QNX only) displays the setting, and
   if you have a port open, it also shows the current open count (with
   C-Kermit's own access always counting as 1).

2.12. The Connection Log

   C-Kermit 7.0 adds the ability to log connections, so you can see where
   you've been and have a record of calls you've made. A connection is
   defined as any communications session that is begun by SET LINE, SET
   PORT, DIAL, SET HOST, TELNET, or RLOGIN. Connections are not logged
   unless you request it; the command is:

   LOG CX [ filename [ { NEW, APPEND } ] ]
          Enables logging of connections in the given file. If the
          trailing { NEW, APPEND } keyword is omitted, the file is opened
          for appending; i.e. new records are written to the end. If NEW
          is specified, a new file is created; if a file of the same name
          already existed, it is overwritten. If the filename is omitted,
          CX.LOG in your home (login) directory is used (note: uppercase).
          To accept all defaults, just use "log connections" (or "l c" for
          short). Synonym: LOG CONNECTIONS.

   CLOSE CX-LOG
          This closes the connection log if it was open. (Note, the CLOSE
          CONNECTION command closes the connection itself).

   SHOW CX
          This shows your current connection, if any, including the
          elapsed time (since you opened it). Synonym: SHOW CONNECTION.

   \v(cx_time)
          This variable shows the elapsed time of your current connection,
          or if there is no current connection, of your most recent
          connection, of if there have been no connections, 0.

   The connection contains one line per connection, of the form:

  yyyymmdd hh:mm:ss username pid p=v [ p=v [ ... ] ]

   where the timestamp (in columns 1-18) shows when the connection was
   made; username is the login identity of the person who made the
   connection; pid is Kermit's process ID when it made the connection. The
   p's are parameters that depend on the type of connection, and the v's
   are their values:

  T = Connection Type (TCP, SERIAL, DIAL, DECNET, etc).
  H = The name of the Host from which the connection was made.
  N = Destination phone Number or Network host name or address.
  D = Serial connections only: Device name.
  O = Dialed calls only: Originating country code & area code if known.
  E = Elapsed time in hh:mm:ss format (or hhh:mm:ss, etc).

   If you always want to keep a connection log, simply add:

  log connections

   to your C-Kermit customization file. Note, however, that if you make a
   lot of connections, your CX.LOG will grow and grow. You can handle this
   by adding a "logrotate" procedure like the following to your
   customization file, before the "log connections" command:

  define LOGROTATE {                    ; Define LOGROTATE macro
      local \%i \%m \%d \%n \%f MAX
      def MAX 4                         ; How many months to keep
      if not def \%1 -                  ; No argument given
        end 1 \%0: No filename given
      if not = 1 \ffiles(\%1) -         ; Exactly 1 file must match
        end 1 \%0: \%1 - File not found
      .\%d := \fsubstr(\fdate(\%1),1,6) ; Arg OK - get file year & month
      if = \%d -                        ; Compare file year & month
        \fsubstr(\v(ndate),1,6) -       ; with current year & month
          end 0                         ; Same year & month - done
      rename \%1 \%1.\%d                ; Different - rename file
      .\%n := \ffiles(\%1.*)            ; How many old files
      if < \%n \m(MAX) end 0            ; Not enough to rotate
      .\%m := \%1.999999                ; Initial compare string
      for \%i 1 \%n 1 {                 ; Loop thru old logs
         .\%f := \fnextfile()           ; Get next file name
         if llt \%f \%m .\%m := \%f     ; If this one older remember it
      }
      delete \%m                        ; Delete the oldest one
  }
  log connections                       ; Now open the (possibly new) log
  logrotate \v(home)CX.LOG              ; Run the LOGROTATE macro

   As you can see, this compares the yyyymm portion of the modification
   date (\fdate()) of the given file (\%1) with the current yyyymm. If
   they differ, the current file has the yyyymm suffix (from its most
   recent modification date) appended to its name. Then we search through
   all other such files, find the oldest one, and delete it. Thus the
   current log, plus the logs from the most recent four months, are kept.
   This is all done automatically every time you start C-Kermit.

   On multiuser systems, it is possible to keep a single, shared,
   system-wide connection log, but this is not recommended since (a) it
   requires you keep a publicly write-accessible logfile (a glaring target
   for mischief), and (b) it would require each user to log to that file
   and not disable logging. A better method for logging connections, in
   UNIX at least, is syslogging (seeckuins.txt Section 15 and
  Section 4.2 of theIKSD Administration Guide for details).

2.13. Automatic Connection-Specific Flow Control Selection

   Beginning in C-Kermit 7.0, the appropriate flow-control method for each
   connection type is kept in a table, for example:

  Remote:           NONE
  Modem:            RTS/CTS
  Direct-Serial:    NONE
  TCPIP:            NONE

   The size of the table and values for each connection type can vary from
   platform to platform. Type "set flow ?" for a list of available
   flow-control types.

   The table is used to automatically select the appropriate kind of flow
   control whenever you make a connection. You can display the table with:

  SHOW FLOW-CONTROL

   The defaults are as follows:

   Remote:
          NONE or XON/XOFF. This is because C-Kermit is not allowed to
          find out what type of connection the incoming user has (*). No
          kind of flow control will work on every kind of connection,
          including (unexpectedly) KEEP, which we would have liked to use,
          but not turning off flow control at the remote end during file
          transfer on TCP/IP connections is fatal to the transfer (except
          in VMS and HP-UX, where it must be set to Xon/Xoff!) Therefore
          if you are dialing in to a serial port on a server (UNIX or VMS)
          where C-Kermit is running, you will need to tell C-Kermit to
          "set flow keep" before transferring files (assuming the modem
          and port are configured correctly by the system administrator;
          otherwise you'll need to give a specific kind of flow control,
          e.g. "set flow xon/xoff"), so in this case C-Kermit will not
          disable flow control, as it must do when you are coming via
          Telnet (directly or through a terminal server, except on VMS and
          HP-UX).

   Modem:
          This applies when you dial out with a modem. In this case, the
          MODEM FLOW-CONTROL setting takes affect after the SET FLOW
          setting, so it can pick the most appropriate flow control for
          the combination of the particular modem and the
          computer/port/driver that is dialing.

   Direct-Serial:
          The default here is NONE because C-Kermit has no way of knowing
          what kind of flow control, if any, is or can be done by the
          device at the other end of the connection. RTS/CTS would be a
          bad choice here, because if the CTS signal is not asserted, the
          connection will hang. And since direct connections are often
          made with 3-wire cables, there is a good chance the CTS signal
          will not be received.

   TCPIP:
          NONE, since TCP and IP provide their own flow control
          transparently to the application, except in VMS, where Xon/Xoff
          is the default due to the requirements of the VMS TCP/IP
          products.

   Other networks:
          NONE, since networks should provide their flow control
          transparently to the application.

   (*) This is possibly the worst feature of UNIX, VMS, and other
   platforms where C-Kermit runs. If C-Kermit was able to ask the
   operating system what kind of connection it had to the user, it could
   set up many things for you automatically.

   You can modify the default-flow-control table with:

  SET FLOW-CONTROL /xxx { NONE, KEEP, RTS/CTS, XON/XOFF, ... }

   where "xxx" is the connection type, e.g.

  SET FLOW /REMOTE NONE
  SET FLOW /DIRECT RTS/CTS

   If you leave out the switch, SET FLOW works as before, choosing the
   flow control method to be used on the current connection:

  SET FLOW XON/XOFF

   Thus, whenever you make a connection with SET PORT, SET LINE, DIAL, SET
   HOST, TELNET, RLOGIN, etc, an appropriate form of flow control is
   selected automatically. You can override the automatic selection with a
   subsequent SET FLOW command, such as SET FLOW NONE (no switch
   included).

   The flow control is changed automatically too when you give a SET MODEM
   TYPE command. For example, suppose your operating system (say Linux)
   supports hardware flow control (RTS/CTS). Now suppose you give the
   following commands:

  set line /dev/ttyS2    ; Automatically sets flow to NONE
  set modem type usr     ; Automatically sets flow to RTS/CTS
  set modem type rolm    ; Doesn't support RTS/CTS so now flow is XON/XOFF

   IMPORTANT: This new feature tends to make the order of SET LINE/HOST
   and SET FLOW commands matter, where it didn't before. For example, in
   VMS:

  SET FLOW KEEP
  SET LINE TTA0:

   the SET LINE undoes the SET FLOW KEEP command; the sequence now must
   be:

  SET FLOW /DIRECT KEEP
  SET LINE TTA0:

   or:

  SET LINE TTA0:
  SET FLOW KEEP

2.14. Trapping Connection Establishment and Loss

   If you define a macro called ON_OPEN, it is executed any time that a
   SET LINE, SET PORT, SET HOST, TELNET, RLOGIN or similar command
   succeeds in opening a connection. The argument is the host or device
   name (as shown by SHOW COMMUNICATIONS, and the same as \v(line)). This
   macro can be used for all sorts of things, like automatically setting
   connection- or host-specific parameters when the connection is opened.
   Example:

  def ON_OPEN {
      switch \%1 {
        :abccorp.com, set reliable off, break
        :xyzcorp.com, set receive packet-length 1000, break
        etc etc...
      }
  }

   If you define a macro called ON_CLOSE, it will be executed any time
   that a SET LINE, SET PORT, SET HOST, TELNET, RLOGIN or any other kind
   of connection that C-Kermit has made is closed, either by the remote or
   by a local CLOSE, HANGUP, or EXIT command or other local action, such
   as when a new connection is opened before an old one was explicitly
   closed.

   As soon as C-Kermit notices the connection has been closed, the
   ON_CLOSE macro is invoked at (a) the top of the command parsing loop,
   or (b) when a connection is closed implicitly by a command such as SET
   LINE that closes any open connection prior to making a new connection,
   or (c) when C-Kermit closes an open connection in the act of exiting.

   The ON_CLOSE macro was inspired by the neverending quest to unite
   Kermit and SSH. In this case using the "tunnel" mechanism:

  def TUNNEL {                                ; \%1 = host to tunnel to
      local \%p
      if not def \%1 stop 1
      assign tunnelhost \%1                   ; Make global copy
      undef on_close
      set macro error off
      close connection                        ; Ignore any error
      open !read tunnel start \%1
      read \%p                                ; Get port number
      if fail stop 1 Tunnel failure: \%1
      close read
      if fail stop 1 Tunnel failure: \%1      ; See Section 4.2.8.1
      assign on_close {                       ; Set up close handler
          echo Closing tunnel: \m(tunnelhost)
          !tunnel stop \m(tunnelhost)
          undef on_close
      }
      set host localhost:\%p /telnet
      if success end 0
      undef on_close
      stop 1 Connection failure: \%1
  }

   In this case, when the connection stops, we also need to shut down the
   tunnel, even if it is at a later time after TUNNEL has finished
   executing. This way we can escape back, reconnect, transfer files, and
   so on until the connection is broken by logging out from the remote, or
   by explicitly closing it, or by EXITing from C-Kermit, at which time
   the tunnel is shut down.

   When the connection is closed, no matter how, the ON_CLOSE macro
   executes and then undefines (destroys) itself, since we don't want to
   be closing tunnels in the future when we close subsequent connections.

   Other such tricks can be imagined, including ending ON_CLOSE with a
   STOP command to force the command stack to be peeled all the way back
   to the top, for example in a deeply nested script that depends on the
   connection being open:

  def on_close { stop 1 CONNECTION LOST }

   When C-Kermit invokes the ON_CLOSE macro, it supplies one argument
   (\%1): the reason the connection was closed as a number, one of the
   following:

  2 - Fatal failure to negotiate a required item on a network connection.
  1 - Closed by C-Kermit command.
  0 - All others (normally closed by remote).

   which may be used for any purpose; for example, to add a comment to the
   connection log:

  def on_close {
      local \%m
      if not open cx end 0
      switch \%1 {
        :0, .\%m = Closed by remote, break
        :1, .\%m = Closed by me, break
        :2, .\%m = Network protocol negotiation failure, break
      }
      if def \%m writeln cx {# \%m}
  }

2.15. Contacting Web Servers with the HTTP Command

   C-Kermit 7.0 (at this writing, the UNIX version only) supports direct
   contact and interaction with Web servers via HTTP 1.0 protocol. To make
   a connection, use Kermit's normal method for making a TCP/IP
   connection, but specify the HTTP port:

  SET HOST host http [ switches ]

   where host is the IP hostname or address, and http is the name of the
   TCP port for the Web server. Relevant switches include:

   /RAW
          Treat the connection as a transparent binary pipe. This switch
          may be required if a port other than 'http' is used.

   /SSL
          Make an secure private connection with SSL (only if SSL support
          is included in your version of Kermit). In this case the port
          name might need to be https rather than http, e.g. "set host
          secureserver.xyzcorp.com https /ssl".

   /TLS
          Make an secure private connection with TLS (only if TLS support
          is included in your version of Kermit). In this case the port
          name would be https rather than http.

   Then you can issue an HTTP command. In most cases, the server closes
   the connection when the command is complete. Example:

  SET HOST www.columbia.edu http
  IF FAIL EXIT 1 Can't contact server
  HTTP GET kermit/index.html

   At this point the connection is closed, since that's how HTTP 1.0
   works. If you want to perform additional operations, you must establish
   a new connection with another SET HOST command.

   The HTTP command acts as a client to the Web server, except instead of
   displaying the results like a Web browser would, it stores them. Any
   HTTP command can (but need not) include any or all of the following
   switches:

   /AGENT:user-agent
          Identifies the client to the server; "C-Kermit" or "Kermit-95"
          by default.

   /HEADER:header-line
          Used for specifying any optional headers. A list of headers is
          provided using braces for grouping:

  /HEADER:{{tag:value}{tag:value}...}

          For a listing of valid tag value and value formats seeRFC
          1945: Hypertext Transfer Protocol -- HTTP/1.0. A maximum of
          eight headers may be specified.

   /USER:name
          In case a page requires a username for access.

   /PASSWORD:password
          In case a page requires a password for access.

   /ARRAY:arrayname
          Tells Kermit to store the response headers in the given array,
          one line per element. The array need not be declared in advance.
          Example:

  C-Kermit? http /array:c get kermit/index.html
  C-Kermit? show array c
  Dimension = 9
  1. Date: Fri, 26 Nov 1999 23:12:22 GMT
  2. Server: Apache/1.3.4 (Unix)
  3. Last-Modified: Mon, 06 Sep 1999 22:35:58 GMT
  4. ETag: "bc049-f72-37d441ce"
  5. Accept-Ranges: bytes
  6. Content-Length: 3954
  7. Connection: close
  8. Content-Type: text/html

   As you can see, the header lines are like MIME e-mail header lines:
   identifier, colon, value. The /ARRAY switch is the only method
   available to a script to process the server responses for a POST or PUT
   command.

   The HTTP commands are:

   HTTP [ switches ] GET remote-filename [ local-filename ]
          Retrieves the named file. If a local-filename is given, the file
          is stored locally under that name; otherwise it is stored with
          its own name.

   HTTP [ switches ] HEAD remote-filename local-filename
          Like GET except without actually getting the file; instead it
          gets only the headers, storing them into the given file, whose
          name must be given, one line per header item, as shown above in
          the /ARRAY: switch description.

   HTTP [ switches ] INDEX remote-directory [ local-filename ]
          Retrieves the file listing for the given server directory. NOTE:
          This command is not supported by most Web servers.

   HTTP [ switches ] POST [ /MIME-TYPE:type ] local-file remote-file
          Used to send a response as if it were sent from a form. The data
          to be posted must be read from a file.

   HTTP [ switches ] PUT [ /MIME-TYPE:type ] local-file remote-file
          Uploads a local file to a server file.

   HTTP [ switches ] DELETE remote-filename
          Instructs the server to delete the specified filename.

3. TERMINAL CONNECTION

3.1. CONNECT Command Switches

   The following switches (see Section 1.5) were added to the CONNECT
   command in 7.0:

   /QUIETLY
          Don't print the "Connecting to..." or "Back at..." messages. CQ
          is an invisible command synonym for CONNECT /QUIETLY.

   /TRIGGER:string
          Specify a trigger or triggers ( Section 3.2) effective for
          this CONNECT command only, temporarily overriding any current
          SET TERMINAL TRIGGER values ( Section 3.2).

   Note: Other switches might also be available; type "connect ?" for a
   list, "help connect" for a description of each.

3.2. Triggers

   Triggers were added for UNIX, VMS, AOS/VS, and K95 in C-Kermit 7.0.

   SET TERMINAL TRIGGER string
          Tells C-Kermit to look for the given string during all
          subsequent CONNECT sessions, and if seen, to return to command
          mode automatically, as if you had escaped back manually. If the
          string includes any spaces, you must enclose it in braces.
          Example:

  set terminal trigger {NO CARRIER}

   Comparisons are made after character-set translation.

   If a string is to include a literal brace character, precede it with a
   backslash:

  ; My modem always makes this noise when the connection is lost:
  set terminal trigger |||ppp\{\{\{\{UUUUUUU

   If you want Kermit to look for more than one string simultaneously, use
   the following syntax:

  set terminal trigger {{string1}{string2}...{stringn}}

   In this case, C-Kermit will return to command mode automatically if any
   of the given strings is encountered. Up to 8 strings may be specified.

   If the most recent return to command mode was caused by a trigger, the
   new variable, \v(trigger), shows the trigger value; otherwise
   \v(trigger) is empty.

   The SHOW TRIGGER command displays the SET TERMINAL TRIGGER values as
   well as the \v(trigger) value.

3.3. Transparent Printing

   As noted in the manual, C-Kermit's CONNECT command on UNIX is not a
   terminal emulator, but rather a "semitransparent pipe" between the
   terminal or emulator you are using to access C-Kermit, and the remote
   host to which C-Kermit is connected. The "semitransparent" qualifier is
   because of character-set translation as well as several actions taken
   by the emulator in response to the characters or strings that pass
   through it, such as APCs, Kermit packets (autodownload), triggers, etc.

   The UNIX version of C-Kermit 7.0 adds another such action: Transparent
   printing, also called Controller printing (as distinct from Autoprint
   or line or screen print). It is intended mainly for use on UNIX
   workstation consoles (as opposed to remote logins), but with some care
   can also be used when accessing C-Kermit remotely.

   Transparent printing is related to APC by sharing C-Kermit's built-in
   ANSI escape-sequence parser to detect "printer on" and "printer off"
   sequences from the host. When the printer-on sequence is received, all
   subsequent arriving characters -- including NUL, control characters,
   and escape sequences -- are sent to the SET PRINTER device instead of
   to your screen until the printer-off sequence is received, or you
   escape back, whichever happens first. These bytes are not translated or
   modified or filtered in any way by Kermit (except for possibly
   stripping of the 8th bit, as noted below), but if filtering or
   translation is desired, this can be accomplished by your SET PRINTER
   selection (e.g. by choosing a pipeline of filters).

   By default, your SET PRINTER device is your default UNIX printer, but
   it can also be a file, a command, or the null device (which causes all
   printer material to be discarded). SeeUsing C-Kermit, 2nd Ed.,
   p.41 for details.

   Transparent printing is controlled by the command:

   SET TERMINAL PRINT { ON, OFF }
          When ON, transparent-print sequences are obeyed, and printing
          occurs on the system where C-Kermit is running. When OFF,
          transparent print sequences are ignored and passed through to
          your actual terminal or emulator, along with the data they
          enclose. OFF is the default, for compatibility with earlier
          C-Kermit releases. As noted in the manual, when the current SET
          PRINTER device is a file, transparent-print material is appended
          to it; the file is not overwritten.

   SET TERMINAL BYTESIZE { 7, 8 }
   SET PARITY { EVEN, ODD, MARK, SPACE, NONE }
          If the terminal bytesize is 7, or PARITY is not NONE, the 8th
          bit of each byte is stripped prior to printing.

   The transparent-print escape sequences are:

   <ESC>[5i
          Printer On. Send all subsequent incoming bytes to the printer
          without any kind of filtering, translation, or alteration. Note:
          <ESC> stands for ASCII character number 27 (decimal), Escape.

   <ESC>[4i
          Printer Off. Resume displaying incoming bytes on the screen.

   These are the same sequences used by DEC VT100 and higher terminals and
   other ANSI X3.64 and ISO 6429 compatible terminals. There is no
   provision for selecting other printer-control sequences.

   Restrictions:

    1. You must SET TERM TRANSPARENT-PRINT ON before you can use this
       feature.
    2. Only the 7-bit forms of the escape sequences are supported. The
       8-bit CSI C1 control is not recognized.
    3. Autoprint is not supported, since this requires a full-fledged
       terminal emulator with direct access to the screen.
    4. The start-print and stop-print sequences pass through to the screen
       (there is no way to avoid this without causing unacceptable delays
       or deadlocks in CONNECT mode). Thus if your terminal or emulator
       also supports transparent printing via these same sequences, an
       empty file will be sent to its printer. Normally this has no
       effect.

   Point (4) is similar to the situation with autodownload and APC -- when
   you have several Kermit clients in a chain, you should take care that
   these features are enabled in only one of them.

   Example 1:

  set printer {|lpr -Plaser}  ; Specify the printer (if not default).
  set term print on           ; Enable transparent printing.
  set term byte 8             ; Enable 8-bit characters.
  connect                     ; Enter CONNECT mode.

   Example 2:

  set printer /home/users/olga/printer.log  ; Send printer material to a file.

   Example 3:

  set printer {| grep -v ^Received | lpr}   ; Filter out some lines

   Then use "pcprint" or "vtprint" commands on the host to initiate
   transparent print operations. SeeUsing C-Kermit, 2nd Ed., p.406
   for details.

   Here is a sample "pcprint" shell script for UNIX:

  #!/bin/sh
  echo -n '<ESC>[5i'
  if [ $# -eq 0 ]; then
    cat
  else
    cat $*
  fi
  echo -n '<FF><ESC>[4i'
  # (end)

   (Replace "<ESC>" by the actual ASCII Escape character and "<FF>" by the
   ASCII Formfeed character).

   If you always want transparent printing enabled, put "set term print
   on" in your C-Kermit customization file (~/.mykermrc in UNIX). The "set
   term bytesize" selection, however, is a property of each separate
   connection.

3.4. Binary and Text Session Logs

   C-Kermit 7.0 corrects an oversight in earlier releases, in which binary
   session logs (SET SESSION-LOG BINARY) translated character sets and
   performed various formatting transformations (e.g. "newline mode")
   before writing characters to the session log. In C-Kermit 7.0,
   binary-mode session logging writes characters as they come in, before
   anything (other that parity-bit stripping) is done to them. Text-mode
   session logging records the characters after processing.

4. FILE TRANSFER

   Every file is transferred either in text mode (which implies
   record-format and character-set translation) or binary mode (in which
   each byte is sent literally without any kind of conversion). The mode
   in which a file is transferred is controlled by (a) the default mode,
   in the absence of any other indications; (b) the SET FILE TYPE command;
   (c) various automatic mechanisms based on client/server negotiations,
   directory information or filename patterns, etc.

   The default FILE TYPE was changed from TEXT to BINARY in C-Kermit 7.0
   because:

     * Transferring a text file in binary mode does less damage than
       transferring a binary file in text mode.
     * Only binary-mode transfers can be recovered from the point of
       failure.
     * The automatic transfer-mode mechanisms switch to text mode on a
       per-file basis anyway, so only those files that are not covered by
       the automatic mechanisms are affected.
     * All file transfers on the Web are done in binary mode, so people
       are accustomed to it and expect it.

4.0. BUG FIXES, MINOR CHANGES, AND CLARIFICATIONS

4.0.0. Filenames with Spaces

   Filenames that contain spaces are a major nuisance to a program like
   Kermit, whose command language is line- and word-oriented, in which
   words are separated by spaces and a filename is assumed to be a "word".
   In general (unless noted otherwise in the description of a particular
   command), there is only one way to refer to such files in Kermit
   commands, and that is to enclose the name in braces:

  send {this file}

   Tells Kermit to send the file whose name is "this file" (two words, no
   quotes). Of course, various circumlocutions are also possible, such as:

  define \%a this file
  send \%a

   BUT, perhaps contrary to expectation, you can't use "\32" to represent
   the space:

  send this\32file

   does not work. Why? Because the Kermit parser, which must work on many
   operating systems including Windows, has no way of knowing what you
   mean by "this\32file". Do you mean a file whose name is "this file" in
   the current directory? Or do you mean a file whose name is "32file" in
   the "this" subdirectory of the current directory? Guessing won't do
   here; Kermit must behave consistently and deterministically in all
   cases on all platforms.

   Note that you can't use Esc or Tab within {...} for filename
   completion, or question mark to get a filename list. However, you can
   include wildcards; for example:

  send {* *}

   sends all files whose name contains a space.

   All things considered, it is best to avoid spaces in file and directory
   names if you can. Also see Section 5.4 on this topic.

4.0.1. Packet out of Window

   C-Kermit 6.0 could send packets "out of window" if the window size was
   greater than 1 and ACKs had arrived out of order. Fixed in 6.1.

4.0.2. MOVE after ADD SEND-LIST

   ADD SEND-LIST followed by MOVE did not delete original files; fixed in
   6.1. Carrier loss was not detected during transfer; in 7.0 C-Kermit
   checks for this (but results can not be guaranteed). In any case, the
   protocol will eventually time out if the connection is lost.

4.0.3. GET and RECEIVE As-Names

   In 5A(190) through 6.0.192, the GET and RECEIVE as-name did not
   properly override the RECEIVE PATHNAMES setting. In 7.0 it does.

4.0.4. New Brief Statistics Listing

   Version 7.0 adds a /BRIEF switch to the STATISTICS command, to display
   a short file-transfer statistics report. /BRIEF is now the default. Use
   /VERBOSE to see the full display, which is about 25 lines long.

4.0.5. Improved FAST Command

   The preinstalled definition of the FAST macro did not take enough
   factors into account. Now it sets packet lengths and window sizes
   appropriate to the configuration. Furthermore, in IRIX only, it might
   restrict the SEND packet length to 4000, to work around a bug in the
   IRIX Telnet server, depending on the IRIX version (seeckubwr.txt,
   IRIX section). To see the built-in definition of the FAST macro, type
   "show macro fast". To change it, simply define it to be whatever you
   want -- it's just a macro, like any other.

4.0.6. The SET SEND BACKUP Command

   Version 7.0 adds SET SEND BACKUP { ON, OFF }. This tells whether backup
   files should be sent. Backup files are the ones created by Kermit (and
   EMACS, and possibly other applications) to preserve old copies of files
   when creating new ones with the same name. Kermit does this when
   receiving a file and its FILE COLLISION setting is BACKUP (or RENAME,
   in which case it the new file gets the backup name). On most platforms,
   the backup name is formed by adding:

  .~n~

   to the end of the filename, where "n" is a number. For example, if the
   original file is oofa.txt, a backup file might be called:

  oofa.txt.~1~

   (or oofa.txt.~2~, etc). If you SET SEND BACKUP OFF, this tells Kermit
   not to send files that have backup names. Normally, SET SEND BACKUP is
   ON (as shown by SHOW PROTOCOL), and backup files are sent if their
   names match the SEND file specification.

   Also see PURGE, SET FILE COLLISION, SEND /NOBACKUP, DIRECTORY
   /[NO]BACKUP.

4.0.7. The SET { SEND, RECEIVE } VERSION-NUMBERS Command

   VMS Only. Normally when sending files, VMS C-Kermit strips the version
   number. For example, if the file is FOO.BAR;34, the name is sent as
   FOO.BAR (without the ";34"). If you want to keep version numbers on
   when sending files, use SET SEND VERSION-NUMBERS ON. The effect depends
   on the receiver.

   Normally when receiving files, and an incoming filename includes a
   VMS-style version number (such as FOO.BAR;34) VMS C-Kermit strips it
   before trying to create the new file; this way the new file receives
   the next highest version number in the customary manner for VMS. If you
   want version numbers on incoming filenames to be used in creating the
   new files, use SET RECEIVE VERSION-NUMBERS ON.

   Normally these commands would be effective only when VMS C-Kermit is
   exchanging files with a non-VMS Kermit program, since VMS-to-VMS
   transfers use labeled mode unless you have gone out of your way to
   defeat it.

   Example: You want to send all versions of all files in the current
   directory from a VMS C-Kermit client to a UNIX C-Kermit server. Use:

  set send version-numbers on
  send *.*;*

   The resulting Unix files will have VMS-style version numbers as part of
   their name, for example "foo.bar;1", "foo.bar;2", etc.

   Now suppose you want to send these files from Unix to another VMS
   system and preserve the version numbers. Again we have a Unix C-Kermit
   server and VMS C-Kermit client. Give these commands to the client:

  set receive version-numbers on
  get *

4.0.8. The SET { SEND, RECEIVE } { MOVE-TO, RENAME-TO } Commands

   These commands are persistent global versions of the /MOVE-TO: and
   /RENAME-TO: switches of the SEND, GET, and RECEIVE commands. They
   should normally be used only when setting up a dedicated
   transaction-processing application, in which each file is to be moved
   or renamed immediately after, and only if, it is transferred
   successfully, so that (for example) an independent, concurrent process
   can notice when new files appear and process them immediately without
   having to guess whether they are complete.

4.0.9. SET FILE INCOMPLETE AUTO

   SET FILE INCOMPLETE { KEEP, DISCARD }, which tells whether to keep or
   discard incompletely received files, has a new option, AUTO, which is
   also the default. It means KEEP the incomplete file if the transfer is
   in binary mode, otherwise DISCARD it. This reduces the chances that a
   subsequent recovery operation (RESEND, REGET, etc) could produce a
   corrupt file, since recovery works only for binary-mode transfers.

4.1. FILE-TRANSFER FILENAME TEMPLATES

   File-transfer filename templates allow files to be renamed
   automatically by the file sender, the receiver, or both, during
   transfer of groups of files.

4.1.1. Templates in the As-Name

   Prior to C-Kermit 6.1 and Kermit 95 1.1.12 the only options that could
   be used to affect the names of files being transferred were SET
   FILENAMES { LITERAL, CONVERTED } and SET { SEND, RECEIVE } PATHNAMES {
   ON, OFF }, plus the "as-name" feature of the SEND (MOVE, etc) and
   RECEIVE commands.

   Previously, the as-name could be used only for a single file. For
   example:

  SEND FOO BAR

   would send the file FOO under the name BAR, but:

  SEND *.TXT anything

   was not allowed, since it would give the same name to each file that
   was sent. When receiving:

  RECEIVE FOO

   would rename the first incoming file to FOO before storing it on the
   disk, but subsequent files would not be renamed to FOO, since this
   would result in overwriting the same file repeatedly. Instead, they
   were stored under the names they arrived with.

   Beginning in C-Kermit 6.1 and Kermit 95 1.1.12, it is possible to
   specify as-names in SEND, RECEIVE, and related commands even for file
   groups. This is accomplished by using replacement variables in the
   as-name, along with optional material such character-string functions
   and/or constant strings. An as-name containing replacement variables is
   called a filename template.

   The key to filename templates is the new variable:

  \v(filename)

   During file transfer it is replaced by the name of each file currently
   being transferred (after transfer, it is the name of the last file
   transferred).

   So, for example:

  send *.txt \v(filename).new

   sends each file with its own name, but with ".new" appended to it. Of
   course if the name already contains periods, this could confuse the
   file receiver, so you can also achieve fancier effects with
   constructions like:

  send *.txt \freplace(\v(filename),.,_).new

   which replaces all periods in the original filename by underscores, and
   then appends ".new" to the result. So, for example, oofa.txt would be
   sent as oofa_txt.new.

   Another new variable that is useful in this regard is \v(filenumber),
   which is the ordinal number of the current file in the file group, so
   you can also:

  send *.txt FILE\flpad(\v(filenum),2,0)

   resulting in a series of files called FILE00, FILE01, FILE02, etc. (At
   the end of the transfer, \v(filenum) tells the number of files that
   were transferred).

   If you specify a constant as-name when sending a file group:

  send *.txt thisnameonly

   Kermit complains and asks you to include replacement variables in the
   as-name. You should generally use \v(filename) or \v(filenumber) for
   this purpose, since other variables (with the possible exception of
   date/time related variables) do not change from one file to the next.
   But Kermit accepts any as-name at all that contains any kind of
   variables for file group, even if the variable will not change. So:

  send *.txt \%a

   is accepted, but all files are sent with the same name (the value of
   \%a, if it has one and it is constant). If the variable has no value at
   all, the files are sent under their own names.

   Of course, the value of \%a in the previous example need not be
   constant:

  define \%a FILE\flpad(\v(filenum),2,0)_at_\v(time)
  send *.txt \%a

   The RECEIVE command, when given without an as-name, behaves as always,
   storing all incoming files under the names they arrive with, subject to
   SET FILE NAME and SET RECEIVE PATHNAMES modifications (Section
   4.10).

   However, when an as-name is given in the RECEIVE command, it is applied
   to all incoming files rather than to just the first. If it does not
   contain replacement variables, then the current FILE COLLISION setting
   governs the result. For example:

  receive foo

   will result in incoming files named foo, foo.~1~, foo.~2~, and so on,
   with the default FILE COLLISION setting of BACKUP. If it does contain
   replacement variables, of course they are used.

   When receiving files, the \v(filename) variable refers to the name that
   was received in the incoming file-header packet, BEFORE any processing
   by SET FILE NAMES or SET RECEIVE PATHNAMES. Since the filenames in
   file-header packets are usually in uppercase, you would need to convert
   them explicitly if you want them in lowercase, e.g.:

  receive \flower(\v(filename)).new

4.1.2. Templates on the Command Line

   On the command-line, use templates as shown above as the -a option
   argument, bearing in mind the propensity of UNIX and perhaps other
   shells to treat backslash as a shell escape character. So in UNIX (for
   example):

  kermit -s oofa.* -a x.\\v(filenum)

   By the way, this represents a change from 6.0 and earlier releases in
   which the as-name (-a argument or otherwise) was not evaluated by the
   command parser. Thus, for example, in VMS (where the shell does not
   care about backslashes), it was possible to:

  kermit -s oofa.txt -a c:\tmp\oofa.txt

   Now backslashes in the as-name must be quoted not only for the shell
   (if necessary) but also for Kermit itself:

  kermit -s oofa.txt -a c:\\tmp\\oofa.txt      ; Kermit only
  kermit -s oofa.txt -a c:\\\\tmp\\\\oofa.txt  ; Shell and Kermit

   You can also use the \fliteral() function for this:

  kermit -s oofa.txt -a \fliteral(c:\tmp\oofa.txt)      ; Kermit only
  kermit -s oofa.txt -a \\fliteral(c:\\tmp\\oofa.txt)   ; Shell and Kermit

4.1.3. Post-Transfer Renaming

   Filename templates are now also useful in SET { SEND, RECEIVE }
   RENAME-TO and in the /RENAME-TO: switch, that can be given to the SEND,
   GET, or RECEIVE commands; this is similar to an as-name, but is
   effective on a per-file basis if and only if the file was transferred
   successfully.

   MOVE-TO and RENAME-TO address a requirement commonly stated for
   transaction processing and similar systems. Suppose, for example, a
   central system "X" accepts connections from multiple clients
   simultaneously; a process on X waits for a file to appear and then
   processes the file. This process must have a way of knowing when the
   file has been completely and successfully transferred before it starts
   to process it. This can be accomplished easily using C-Kermit's SET {
   SEND, RECEIVE } { MOVE-TO, RENAME-TO } command or /MOVE-TO: or
   /RENAME-TO: switches, described inSections 4.7.1 through
  4.7.3.

   Here's an example for the client side, in which files to be sent are
   placed in a certain directory (/usr/olga/tosend in this example) by
   another process when they are ready to go. This might be in a hospital
   or big doctor's office, where medical insurance claims are entered at a
   number of workstations, and then deposited in the "tosend" directory,
   from which they are sent to a claims clearinghouse. We assume the
   connection is already made and a Kermit server is on the other end.

  local srcdir findir              ; Declare local (automatic) variables
  assign srcdir /usr/olga/tosend   ; Local source directory (files to send)
  assign findir /usr/olga/sent     ; Where to move files after they are sent
  log transactions                 ; Keep a log of transfers
  cd \m(srcdir)                    ; Change to the source directory
  while true {                     ; Loop forever...
      send /move-to:\m(findir) *   ; Send all files
      sleep 60                     ; Sleep a minute
  }                                ; Go back and do it again

   Note how simple this is. Once each file is sent, it is moved so it
   won't be sent again (you could also use SEND /RENAME-TO: or even SEND
   /DELETE). If a transfer fails, the file is not moved and so we try
   again to send it next time around. If there are no files to send, the
   SEND command does nothing but a message is printed; you can avoid the
   message by checking first to see if any files are in the directory:

  while true {                     ; Loop forever...
      if > \ffiles(*) 0 -          ; If there are any files
        send /move-to:\m(findir) * ; send them.
      sleep 60                     ; Sleep a minute.
  }                                ; Go back and do it again.

   It's even simpler on the server side (here again we assume the
   connection is already in place):

  local rcvdir findir              ; Declare local (automatic) variables
  assign rcvdir /usr/ivan/tmp      ; Temporary receiving directory
  assign findir /usr/ivan/new      ; Where to move files after reception
  log transactions                 ; Keep a log of transfers
  cd \m(rcvdir)                    ; Change to the source directory
  set receive move-to \m(findir)   ; Declare move-to directory.
  server                           ; Enter server mode.

   A separate process (e.g. the medical claim-form decoder) can look for
   files appearing in the /usr/ivan/new directory and process them with
   every confidence that they have been completely received.

   Note that the use of MOVE-TO can result in moved files overwriting one
   another (the application would normally avoid this by assigning each
   transaction a unique, e.g. based on customer number and claim number).
   But if filename collisions are a possibility in your application,
   RENAME-TO might be a better choice; you can use any variables you like
   in the template to ensure uniqueness of the RENAME-TO filename; for
   example:

  SET RECEIVE RENAME-TO \v(filename)_\v(ndate)_\v(ntime)_\v(userid)_\v(pid)

4.2. FILE-TRANSFER PIPES AND FILTERS

4.2.1. INTRODUCTION

   Beginning in C-Kermit 6.1 and Kermit 95 1.1.12, it is possible to send
   from a command, or "pipe", as well as from a file, and to receive to a
   pipe or command. In a typical example, we might want to transfer an
   entire directory tree from one UNIX system to another (but without
   using the methods described inSections 4.3 ,4.10,
  4.11, and4.15). We could do this in multiple steps as
   follows:

  1. Create a tar archive of the desired directory tree
  2. Compress the tar archive
  3. Transfer it in binary mode to the other computer
  4. Decompress it
  5. Extract the directory tree from the tar archive

   But this is inconvenient and it requires a temporary file, which might
   be larger than we have room for.

   The new pipe-transfer feature lets you do such things in a single step,
   and without intermediate files.

   Additional new features, also discussed here, let you specify pre- and
   post- processing filters for outbound and incoming files, and give you
   a way to insert the output from shell or system commands into C-Kermit
   commands.

   The file-transfer related features are available only with Kermit
   protocol, not with any external protocols, nor with K95's built-in
   XYZMODEM protocols (because XYZMODEM recovers from transmission errors
   by rewinding the source file, and you can't rewind a pipe).

   This section begins by discussing the simple and straightforward use of
   these features in UNIX, in which pipes and input/output redirection are
   a fundamental component and therefore "just work", and then goes on to
   discuss their operation in Windows and OS/2, where matters are much
   more complicated.

4.2.1.1. TERMINOLOGY

   Standard Input
          This is a precise technical term denoting the normal source of
          input for a command or program, which is the keyboard of your
          terminal by default, but which can be redirected to a file or
          pipe.

   Stdin
          Abbreviation for Standard Input.

   Standard Output
          A precise technical term denoting the normal destination for
          output from a command or program, which is your terminal screen
          by default, but which can be redirected to a file.

   Stdout
          Abbreviation for Standard Output.

   Stdio
          Abbreviation for Standard Input / Standard Output.

   I/O
          Abbreviation for Input / Output.

   Shell
          Text-based system command processor, such as the UNIX shell, DOS
          COMMAND.COM, etc.

   Pipe
          A mechanism by which the standard output of one program is sent
          to the standard input of another.

   Pipeline
          A series of programs connected by pipes.

4.2.1.2. NOTATION

   In command descriptions, "command" is replaced by a shell or system
   command or pipeline. The command names specified in these commands are
   interpreted by your shell, just as if you were typing them at the shell
   prompt, and so if they are in your PATH, they will be found in the
   expected manner. Therefore you don't have to specify complete pathnames
   for commands that are programs (but it shouldn't hurt if you do).

   The normal notation for I/O redirection is as follows:

  <  Read Stdin from the given file.
  >  Send Stdout to the given file.
  |  Send Stdout from the command on the left to the command on the right.

   Examples:

   sort < foo > bar
          Sorts the lines in file "foo" and writes the results to file
          "bar"

   grep -c "some text" *.txt | grep -v ":0" | sort | pr -3 | lpr
          This is a command pipeline composed of 5 commands:

   grep -c "some text" *.txt
          Looks in all files whose names end with ".txt" for the string
          "some text" and writes to Stdout the names of each file followed
          by a colon and the number of occurrences in each.

   grep -v ":0"
          Prints to Stdout the lines from Stdin that do NOT contain the
          string ":0", in this case, it removes the names of files that do
          not contain "some text".

   sort
          Sorts the lines from Stdin alphabetically to Stdout.

   pr -3
          Arranges the lines from Stdin in three columns.

   lpr
          Prints its Stdin on the default printer.

   Note that the Kermit features described here work only with commands
   that use Stdio. If you attempt to use them with commands whose input
   and output can not be redirected, Kermit will most likely get stuck.
   Kermit has no way of telling how an external command works, nor what
   the syntax of the shell is, so it's up to you to make sure you use
   these features only with redirectable commands.

   The quoting rules of your shell apply to the command. Thus in UNIX,
   where C-Kermit tries to use your preferred shell for running commands,
   shell "metacharacters" within commands must be escaped if they are to
   be taken literally, using the methods normal for your shell. For
   example, the UNIX tr (translate) command must have its arguments in
   quotes:

  tr "[a-z]" "[A-Z]"

   otherwise the shell is likely to replace them by all filenames that
   match, which is probably not what you want. This is also true when
   using your shell directly, and has nothing to do with Kermit.

4.2.1.3. SECURITY

   Some sites might not wish to allow access to system commands or
   external programs from within Kermit. Such access, including all the
   features described here, can be disabled in various ways:

    1. When building from source code, include -DNOPUSH among the CFLAGS.
    2. At runtime, give the NOPUSH command.
    3. For server mode, give the DISABLE HOST command.
    4. Implicit use of pipes can be disabled as described inSection
       4.2.4.

   Note: 3 and 4 are not necessary if you have done 1 or 2.

4.2.2. Commands for Transferring from and to Pipes

   SEND /COMMAND sends data from a command or command pipeline, and
   RECEIVE /COMMENT writes data to a command or pipeline. The GET /COMMAND
   command asks a server to send material, and then writes the incoming
   material to a command or pipeline. These features, along with switches
   (like "/COMMAND", described in Section 4.7) are new to C-Kermit
   6.1. The following synonyms are also provided:

  CSEND    = SEND /COMMAND
  CRECEIVE = RECEIVE /COMMAND
  CGET     = GET /COMMAND

   None of these commands can be used if a SEND or RECEIVE FILTER
   (respectively, Section 4.2.3) is in effect, or if a NOPUSH command
   ( Section 4.2.1.3) has been given, or if the current protocol is
   not Kermit.

4.2.2.1. Sending from a Command

   SEND /COMMAND command [ as-name ]
   SEND /AS-NAME:as-name /COMMAND command
   CSEND command [ as-name ]
          These three forms are the same. They work like the SEND command,
          but instead of sending a file, it sends the standard output of
          the given command, either under the command's own name, or else
          with the given as-name. If the command contains spaces, it must
          be enclosed in braces. Braces should also be used for the
          as-name if it contains spaces. If braces are included around
          either the command or the as-name, they are removed after
          parsing but before use. As with SEND, the transfer is in text or
          binary mode according the current FILE TYPE setting, unless you
          override the global transfer mode by including a /TEXT or
          /BINARY switch. The command must require no input.

   When sending from a command or pipeline, C-Kermit has no way of knowing
   in advance how much data will be sent, and so it can not send the size
   to the other Kermit in the Attribute packet, and so the receiving
   Kermit has no way of displaying "percent done" or a progress bar
   (thermometer).

   Examples that make sense in text mode (illustrated by common UNIX
   commands):

   SEND /COMMAND finger
   CSEND finger
          sends the current "finger" listing (who's logged in) under the
          name "finger". The two forms "send /command" and "csend" are
          equivalent; we won't bother showing them both in the rest of the
          examples.

   SEND /COMMAND:{finger}
   CSEND {finger}
          Same as previous example (braces are removed from "{finger}").

   SEND /COMMAND:{ finger }
   CSEND { finger }
          Same as previous example, but note that the spaces are kept.
          This does not prevent the shell from running the "finger"
          program, but its output is sent under the name " finger " (with
          a leading and trailing space).

   SEND /COMMAND:finger /AS-NAME:userlist
   CSEND finger userlist
          sends the current finger listing under the name "userlist".

   SEND /COMMAND:{finger | sort -r} /AS-NAME:userlist
   CSEND {finger | sort -r} userlist
          sends the current finger listing, sorted in reverse order, under
          the name "userlist". The braces are needed to distinguish the
          command from the as-name.

   SEND /COMMAND:{finger | sort -r} /AS-NAME:{userlist}
   CSEND {finger | sort -r} {userlist}
          Same as previous example (braces are removed from "{userlist}").

   SEND /COMMAND:{finger | sort -r}
          /AS-NAME:{\freplace(\v(filename),\32,_)}

   CSEND {finger | sort -r} {\freplace(\v(filename),\32,_)}
          Like the previous example, but sends the output of the command
          under the name of the command, but with all spaces (\32)
          replaced by underscores, so the as-name is "finger_|_sort_-r".

   Examples that make sense in binary mode (three equivalent forms are
   shown):

   SEND /COMMAND /BINARY {tar cf - . | gzip -c} mydir.tar.gz
   SEND /COMMAND /BINARY /AS-NAME:mydir.tar.gz {tar cf - . | gzip -c}
   CSEND /BINARY {tar cf - . | gzip -c} mydir.tar.gz
          Makes a tar archive of the current directory, compresses it with
          the GNU gzip program, and sends it as "mydir.tar.gz". The other
          Kermit can, of course, just store it as a file, or it can use
          CRECEIVE to uncompress and dearchive it as part of the transfer
          process.

   When using a "pipeline" of commands in the command field, obviously,
   the first command must not require any input, and the last command
   should produce some output, and all intermediate commands should get
   some input and produce some output.

4.2.2.2. Receiving to a Command

   RECEIVE /COMMAND command
   CRECEIVE command
          This is like RECEIVE, except incoming material is written to the
          standard input of the given command, in text or binary mode
          according to the normal rules for file reception. Be sure to
          include a redirector to a file (if the command normally writes
          to standard output), or the output of the command won't go
          anywhere. The command may contain spaces; braces are not needed,
          but they are removed if used.

   WARNING: C-Kermit has no way of knowing anything about the command, or
   even whether it is a command. Thus this command will always cause
   C-Kermit to enter protocol mode, as long as some text is specified in
   the command field. However, if the text does not correspond to a
   command, the transfer will eventually fail with a message such as
   "Error writing data" or "Failure to close file".

   Examples for text mode (in UNIX):

   RECEIVE /COMMAND sort -r > reverse.txt
   CRECEIVE sort -r > reverse.txt
          The text that is received is sorted in reverse order and stored
          in the file "reverse.txt". The two forms shown are equivalent.

   RECEIVE /COMMAND {sort -r > reverse.txt}
   CRECEIVE {sort -r > reverse.txt}
          The same as the previous example; if braces are included, they
          are simply removed.

   RECEIVE /COMMAND {sort -r > \flower(\v(filename)).reverse}
   CRECEIVE {sort -r > \flower(\v(filename)).reverse}
          Same but stores result under the incoming filename, lowercased,
          and with ".reverse" appended to it.

   RECEIVE /COMMAND sort
   CRECEIVE sort
          Does nothing useful, since the output of sort has nowhere to go.

   RECEIVE /COMMAND sort -r | pr -3 | lpr -Plaserjet
   CRECEIVE sort -r | pr -3 | lpr -Plaserjet
          The text that is received is sorted in reverse order, arranged
          into three columns, and sent to the "laserjet" printer.

   Examples for binary mode:

   RECEIVE /COMMAND:{gunzip -c | tar xf -}
   CRECEIVE {gunzip -c | tar xf -}
          Assuming the data that is received is a compressed tar archive,
          uncompresses the archive and passes it to tar for extraction. In
          this case the braces are needed because otherwise the final "-"
          would be taken as a command continuation character (see
         Using C-Kermit, 2nd Edition, p.33).

   GET /COMMAND remote-file command
   GET /COMMAND /AS-NAME:command remote-file
   CGET remote-file command
          This command tells the Kermit client to send a GET request for
          the given remote file to a Kermit server. Unlike GET, however,
          the incoming material is written to a command, rather than to a
          file. If the remote-file or the command contain spaces, they
          must be enclosed in braces. The same cautions about the command
          apply as for CRECEIVE.

   Examples (for UNIX):

   GET /COMMAND oofa.txt sort -r > oofa.new
   GET /COMMAND {oofa.txt} {sort -r > oofa.new}
   CGET oofa.txt sort -r > oofa.new
   CGET {oofa.txt} {sort -r > oofa.new}
          These four are equivalent. Each of them requests the server to
          send its "oofa.txt" file, and as it arrives, it is sorted in
          reverse order and written to "oofa.new".

   GET /COMMAND {profile exec a} lpr
   GET /COMMAND {profile exec a} {lpr}
   GET /COMMAND /AS-NAME:lpr {profile exec a}
   GET /COMMAND /AS-NAME:{lpr} {profile exec a}
   GET /COMMAND /AS:lpr {profile exec a}
   CGET {profile exec a} lpr
   CGET {profile exec a} {lpr}
          Here the remote filename contains spaces so it MUST be enclosed
          in braces. As it arrives it is sent to the lpr program for
          printing. Braces are optional around "lpr" since it contains no
          spaces.

   GET /COMMAND *.txt {cat >> new.txt}
   GET /AS-NAME:{cat >> new.txt} /COMMAND *.txt
   CGET *.txt {cat >> new.txt}
          This gets all the ".txt" files from the server and concatenates
          them all into a single "new.txt" file on the client.

   GET /COMMAND *.txt {echo \v(filename)>>new.txt;cat>>new.txt}
   CGET *.txt {echo \v(filename)>>new.txt;cat>>new.txt}
          As above, but inserts each file's name before its contents.

4.2.3. Using File-Transfer Filters

   The commands described in Section 4.2.2 let you send the output of
   a command, or receive data into a command. But what if you want to
   specify preprocessing for files that you send, or postprocessing of
   files that you receive, even when multiple files are involved? For this
   you need a way to specify send and receive filters. The commands are
   SET SEND FILTER and SET RECEIVE FILTER; SHOW PROTOCOL displays the
   current settings.

4.2.3.1. The SEND Filter

   SET SEND FILTER [ command ]
          This command specifies a command to be run on any file that you
          SEND (or MOVE, MSEND, etc). It also applies to files sent when
          in server mode, in response to GET commands, but not to the
          results of REMOTE commands like REMOTE DIRECTORY, REMOTE TYPE,
          REMOTE HOST, etc. The command may be, but need not be, enclosed
          in braces; if it is, the braces are stripped before use. The
          output of this command is sent, rather than the file itself. The
          current FILE TYPE setting (TEXT or BINARY) applies to the output
          of the command. The command must contain at least one instance
          of \v(filename), for which the name of the actual file is
          substituted. If the command is omitted, the send filter is
          removed and files are sent in the normal manner.

   The SET SEND FILTER sets up a "global" filter -- that is, one that
   applies to all subsequent file-sending commands until you change or
   remove it. You can also specify a "local" filter to be used in a
   specific file-sending command by using the /FILTER switch (see
  Section 1.5); for example:

  SEND /FILTER:command [ other-switches ] filename

   Besides \v(filename), you can include any other script programming
   notation in the send filter: variable names, array references, calls to
   built-in string or other functions, and so on. These are evaluated
   during file transfer, NOT during parsing, and they are evaluated
   separately for each file.

   When the SEND or MOVE (SEND /DELETE) command is used with a send
   filter, the output from the filter is sent under the file's original
   name unless you specify an "as-name" or template. The Attribute packet
   (if any) contains the original file's attributes (size, creation date,
   etc). So (for example) if the filter changes the file's size, the
   progress thermometer might be wrong. (We can't send the size of the
   output from the filter, because it is not known until the transfer is
   finished.) If you prefer that the size not be sent, use "set attributes
   size off".

   You can not use send filters with RESEND (SEND /RECOVER) or PSEND (SEND
   /START).

   Examples for text mode:

   SET SEND FILTER sort -r \v(filename) ; Braces may be omitted
   SET SEND FILTER {sort -r \v(filename)} ; Braces may be included
   SEND *.txt
          This sends every file in the current directory whose name ends
          with ".txt" under its own name, but with its lines sorted in
          reverse order.

   SEND /FILTER:{sort -r \v(filename)} *.txt
          Same as above, but the filter applies only to this SEND command.
          Braces are required in this case.

   SET SEND FILTER {sort -r \v(filename)}
   SEND oofa.txt reverse.txt
          Sends the oofa.txt file with its lines sorted in reverse order
          under the name "reverse.txt".

   SET SEND FILTER {sort -r \v(filename)}
   SEND oofa.* \v(filename).reverse
          Sends all the oofa.* files with their lines sorted in reverse
          order; each file is sent under its own name but with ".reverse"
          appended to it.

   SET SEND FILTER {tr "[a-z]" "[A-Z]" < \v(filename)}
   SEND *.txt
          Sends all ".txt" files under their own names, but uppercasing
          their contents.

   Note that the SEND FILTER applies not only to files that are sent with
   SEND, MOVE, MSEND, etc, but also to files sent by the C-Kermit server
   in response to GET requests.

   Examples for binary mode:

   SET SEND FILTER {gzip -c \v(filename)}
   SEND /BINARY oofa.txt oofa.txt.gz
          Sends the oofa.txt file, compressed by gzip, as oofa.txt.gz.

   SEND /BINARY /FILTER:{gzip -c \v(filename)} oofa.txt oofa.txt.gz
          As above, but the filter applies only to this SEND command.

   SET SEND FILTER {gzip -c \v(filename)}
   SEND /BINARY oofa.* \fupper(\replace(\v(filename),.,_)).GZ
          Sends all the oofa.* files, compressed by gzip, each under its
          own name, but with the name uppercased, all periods within the
          name converted to underscores, and ".GZ" appended to it. So, for
          example, "oofa.txt" is sent as "OOFA_TXT.GZ".

   In the gzip examples, note that the amount of data that is sent is
   normally less than the original file size because gzip compresses the
   file. But Kermit sends the original file size ahead in the attribute
   packet anyway (unless you tell it not too). Thus the transfer will
   probably appear to terminate early, e.g. when the receiver's
   file-transfer display thermometer is only at 40%. If this annoys you,
   tell Kermit to "set attribute length off". On the other hand, you can
   use the final position of the thermometer as a measure of the
   effectiveness of compression.

4.2.3.2. The RECEIVE Filter

   SET RECEIVE FILTER [ command ]
          This command specifies that the given command will be run on any
          file that is received before it is written to disk. The command
          may be, but need not be, enclosed in braces; if it is the braces
          are stripped before use. The following two commands are
          equivalent:

  SET RECEIVE FILTER sort -r > \v(filename)
  SET RECEIVE FILTER {sort -r > \v(filename)}

   The RECEIVE filter command may contain a "\v(filename)" sequence to be
   replaced by the incoming filename from the file header packet, but it
   is not required. However you must use it whenever your filter would
   normally write to Stdout, otherwise its output will be lost.

   The RECEIVE filter command may contain one or more "\v(filename)"
   sequence to be replaced by the incoming filename from the file header
   packet, but it is not required. However you must use it whenever your
   filter would normally write to Stdout, otherwise its output will be
   lost.

   RECEIVE /FILTER:command and GET /FILTER:command can also be used to
   specify a filter to be used for only one file-transfer operation.

   UNIX examples for text mode:

   SET RECEIVE FILTER lpr
   RECEIVE
          All the files that are received are sent to the default UNIX
          print spooler.

   RECEIVE /FILTER:lpr
          Same as above, except the lpr filter is used only with this
          RECEIVE command.

   RECEIVE lpr
          This is probably not what you want; it creates a file called
          lpr.

   SET RECEIVE FILTER {sort -r > \v(filename)}
   RECEIVE
          Stores each incoming file with its lines sorted in reverse
          order, under its own name.

   RECEIVE /FILTER:{sort -r > \v(filename)}
          As above, but the filter is used only for this RECEIVE command.

   SET RECEIVE FILTER sort -r > \v(filename)
   RECEIVE reverse.txt
          Stores each incoming file with its lines sorted in reverse
          order, under the name "reverse.txt". The actual result depends
          on the FILE COLLISION setting. If it is OVERWRITE and multiple
          files arrive, then each incoming file destroys the previous one.
          If it is BACKUP (the default), filename conflicts are resolve by
          adding "version numbers" to the filenames: reverse.txt,
          reverse.txt.~1~, reverse.txt.~2~, etc.

   SET RECEIVE FILTER sort -r > \v(filename)
   RECEIVE \v(filename).reverse
          Stores each incoming file with its lines sorted in reverse
          order, under the name it arrived with, but with ".reverse"
          appended to it.

   SET RECEIVE FILTER sort -r > \v(filename)
   RECEIVE \flower(\v(filename)).reverse
          Like the previous example, but ensures that the filename is
          lowercase.

          Examples for binary mode:

   SET RECEIVE FILTER gunzip -c > \v(filename)
   RECEIVE
          This receives one or more presumably compressed file and
          uncompresses each one into a file having the same name it was
          sent with. For example, if the file is sent with the name
          OOFA.TXT.GZ, it is stored with that name, even after
          decompression.

   SET RECEIVE FILTER gunzip -c > \v(filename)
   RECEIVE \flower(\fsubstring(\v(filename),1,\flength(\v(filename))-3))
          Like the previous example, but the resulting filename has its
          rightmost three characters removed from it and the remainder is
          lowercased. So if the incoming filename is OOFA.TXT.GZ, it is
          stored as oofa.txt after decompression.

   Of course you don't want to type such long hideous commands, so we have
   also introduced several new functions:

   \fstripx(string[,character])
          This function removes the rightmost segment of the string that
          starts with the given character. If no character is given,
          period (.) is used. Thus it is most conveniently used for
          stripping the extension from a filename (or the decimal portion
          from a floating-point number written in US/UK style). Examples:

   \fstripx(OOFA.TXT.GZ)             => OOFA.TXT
   \fstripx(OOFA.TXT.GZ,.)           => OOFA.TXT
   \fstripx(OOFA.TXT.GZ,X)           => OOFA.T
   \fstripx(\fstripx(OOFA.TXT.GZ))   => OOFA
   \fstripx($100.00)                 => $100

   \fstripn(string,number)
          Removes the rightmost number characters from the string.
          Examples:

   \fstripn(OOFA.TXT.GZ)             => OOFA.TXT.GZ
   \fstripn(OOFA.TXT.GZ,3)           => OOFA.TXT
   \fstripn(OOFA.TXT.GZ,7)           => OOFA

   \fstripb(string[,c1[,c2]])
          Strips enclosing matching braces, brackets, parentheses, or
          quotes from the string. The second argument, c1, specifies which
          kind of enclosure to look for; if not specified, any enclosing
          (), [], <>, {}, "", '', or `' are removed. If c1 is specified
          and c2 is not, then if c1 is an opening brace, bracket, or
          parenthesis, the matching closing one is supplied automatically
          as c2. If both c1 and c2 are specified, then to be stripped the
          string must begin with c1 and end with c2. If the string is not
          enclosed in the indicated manner, the result is the original
          string. Examples:

   \fstripb("abc")                   => abc
   \fstripb([abc])                   => abc
   \fstripb([abc)                    => [abc
   \fstripb(<abc>)                   => abc
   \fstripb(<abc>,[)                 => <abc>
   \fstripb((abc))                   => abc
   \fstripb((abc),[)                 => (abc)
   \fstripb((abc),{(})               => abc
   \fstripb(+abc+)                   => +abc+
   \fstripb(+abc+,+)                 => abc
   \fstripb(+abc+,+,^)               => +abc+
   \fstripb(+abc^,+,^)               => abc
   \fstripb('abc')                   => abc
   \fstripb(`abc')                   => abc
   \fstripb(``abc'')                 => `abc'
   \fstripb(\fstripb(``abc''))       => abc

          Notice the special syntax required for including a literal
          parenthesis in the argument list. As the last two examples
          illustrate, \fstripb() strips only one level at at a time;
          nesting can be used to strip a small fixed number of levels;
          loops can be used to strip larger or indeterminate numbers of
          levels.

   \flop(string[,char])
          Removes the leftmost segment of the string that ends with the
          given character. If no character is given, period (.) is used.
          Examples:

  \flop(OOFA.TXT.GZ)               => TXT.GZ
  \flop(OOFA.TXT.GZ,.)             => TXT.GZ
  \flop(OOFA.TXT.GZ,X)             => T.GZ

          To remove the leftmost number characters, just use
          \fsubstring(s,number+1). To return the rightmost number
          characters, use \fright(s,number).

   So the hideous example:

  receive \flower(\fsubstring(\v(filename),1,\flength(\v(filename))-3))

   can now be written as:

  receive \flower(\fstripx(\v(filename)))

   That is, the filename stripped of its extension and then lowercased.
   This is not only shorter and less hideous, but also does not depend on
   the length of the extension being 3.

   Note that when a receive filter is in effect, this overrides your FILE
   COLLISION setting, since Kermit has no way of knowing what the final
   destination filename will be (because it does not know, and can not be
   expected to know, the syntax of every version of every command shell on
   every platform on the planet).

4.2.4. Implicit Use of Pipes

   If you wish, C-Kermit can also examine incoming filenames to see if
   they start with "!", and if so, the subsequent text is treated as a
   command to read from or write to. For example, if a Kermit client is
   given the following command:

  get {!finger | sort}

   the server on the other end, if it supports this feature, will run the
   "finger" program, pipe its standard output to the "sort" program, and
   send sort's standard output back to you. Similarly, if you:

  send oofa.txt !sort -r > oofa.new

   or, equivalently:

  send oofa.txt {!sort -r > oofa.new}

   or:

  send /as-name:{!sort -r > oofa.new} oofa.txt

   this has the receiver send the contents of the incoming oofa.txt file
   to the sort program, which sorts the text in reverse order and stores
   the result in oofa.new.

   This use of the exclamation mark should be familiar to UNIX users as
   the "bang" feature that lets you run an external application or command
   from within another application.

   Kermit's "bang" feature is disabled by default, since it is not unheard
   for filenames to actually begin with "!". So if you want to use this
   feature, you must enable it with the following command:

   SET TRANSFER PIPES { ON, OFF }
          ON enables the recognition of "!" notation in incoming filenames
          during file transfer as an indicator that the remaining text is
          the name of a command. OFF, the default, disables this feature
          and uses the text as a filename in the normal fashion. This
          command does NOT affect SEND /COMMAND, GET /COMMAND, CSEND, etc.

   So using a combination of CSEND (SEND /COMMAND) and the "bang" feature,
   you can transfer a directory tree all in one command (assuming the
   remote Kermit supports pipe transfers and has them enabled):

  CSEND {tar cf - . | gzip -c} {!gunzip -c | tar xf -}

   or:

  SEND /COMMAND:{tar cf - . | gzip -c} /as:{!gunzip -c | tar xf -}

   Pay close attention to the syntax. Braces are needed around the command
   because it contains spaces; braces are needed around the as-name
   because it ends with "-". The as-name must begin with "!" or receiving
   Kermit will not recognize it as a command. The CSEND command must NOT
   begin with "!" unless you are running a command whose name really does
   start that character.

   Similarly, you have a Kermit server send a directory tree to be
   unpacked on the client end:

  CGET {!tar cf - . | gzip -c} {gunzip -c | tar xf -}

   or:

  GET /COMMAND {!tar cf - . | gzip -c} /as:{gunzip -c | tar xf -}

   Notice how, in this case, the bang is required in the remote command,
   to distinguish it from a filename, but not in the local command, since
   by definition of CGET (or GET /COMMAND), it is known to be a command.

   SEND and RECEIVE FILTERs supersede the bang feature. For example, if a
   file arrives under the name "!gunzip -c | tar xf -", but the receiving
   Kermit also has been given a command like:

  set receive filter sort -r > \v(filename)

   then the incoming data will be sorted rather than gunzipped.

   Finally, if SET TRANSFER PIPES is ON (and in this case, this must be
   done in your C-Kermit initialization file), you can send from a pipe on
   the C-Kermit command line:

  kermit -s "!finger | sort -r" -a userlist

   In this case the "filename" contains spaces and so must be quoting
   using your shell's quoting rules.

4.2.5. Success and Failure of Piped Commands

   Commands or programs started by Kermit as a result of CSEND or CRECEIVE
   commands, CGET, SEND /COMMAND, REDIRECT commands (seeSection
   4.2.8.2), implicit use of pipes, RUN commands, and so forth, should
   return their exit status codes to the Kermit command that caused them
   to be run, and therefore IF SUCCESS and IF FAILURE tests after these
   commands should work as expected. For example:

  CSEND blah < filename

   should fail if there is no command called "blah" or if there is no file
   called "filename". However, this is not foolproof and sometimes
   C-Kermit might think a command succeeded when it failed, or vice versa.
   This is most likely to happen when the highly system-dependent methods
   that Kermit must use to determine a command's exit status code do not
   supply the right information.

   It can also happen because some commands might define success and
   failure differently from what you expect, or because you are using a
   pipeline composed of many commands, and one of them fails to pass
   failing exit status codes up the chain. The most likely culprit is the
   shell itself, which in most cases must be interposed between Kermit and
   any external program to be run.

   In any case, you can examine the following variable to find out the
   exit status code returned to Kermit by the process most recently run by
   any command that runs external commands or programs, including CSEND,
   CRECEIVE, REDIRECT, RUN, etc:

  \v(pexitstat)

   In UNIX, Windows and OS/2, the value should be -2 if no command has
   been run yet, 0 if the most recent command succeeded, -1, -3, or -4 if
   there was an internal error, and a positive number returned by the
   command itself if the command failed. If the number is in the range
   1-127, this is the program's exit status code. If it is 128 or greater,
   this is supposed to indicate that the command or program was
   interrupted or terminated from outside itself.

   In Windows 95 and 98, the return values of the default shell are
   unreliable; various third-party shells can be used to work around this
   deficiency.

   In VMS, it is the actual exit status code of the command that was run.
   This is an odd number if the command succeeded, and an even number if
   it failed. You can see the associated message as follows:

  run write sys$output f$message(\v(pexitstat))

   Or, more conveniently, use the new Kermit function:

  echo \ferrstring(\v(pexitstat))

   which converts a system error code (number) to the corresponding
   message.

4.2.6. Cautions about Using Pipes to Transfer Directory Trees

   Although utilities such as tar and zip/unzip might be available on
   different platforms (such as UNIX and Windows), this does not
   necessarily mean you can use them successfully to transfer directory
   trees between unlike platforms. For example:

  CSEND {tar cf - . | gzip -c} {!gunzip -c | tar xf -}

   when used from UNIX to Windows will have satisfactory results for
   binary files, but not for text files. UNIX text files have lines ending
   with Linefeed (LF) only, whereas Windows text files have lines ending
   in Carriage Return and Linefeed (CRLF). Thus any text files that were
   in the archive formed by the first tar command will be unpacked by the
   second tar command in their original form, and will display and print
   incorrectly in Windows (except in applications that have been
   explicitly coded to handle UNIX-format text files). On the other hand
   if you told gzip to use "text mode" to do record format conversion
   (assuming there was a way to tell it, as there is with most "zip"
   programs), this would destroy any binary files in the archive.

   Furthermore, if the archive contains text files that are written in
   languages other than English, the "special" (accented and/or non-Roman)
   characters are NOT translated, and are therefore likely show up as
   gibberish on the target system. For example, West European languages
   are usually encoded in ISO Latin Alphabet 1 in UNIX, but in PC code
   page 850 on the PC. Capital A with acute accent is code point 193
   (decimal) Latin-1, but 181 in CP850. So A-acute in the UNIX file
   becomes Middle Box Bottom on the PC, and similarly for all the other
   special characters, and for all other languages -- Greek, Russian,
   Hebrew, Japanese, etc.

   So when transferring text files between unlike platforms, you should
   use direct Kermit file transfers so Kermit can apply the needed
   record-format and character-set transformations. Use pipelines
   containing archivers like tar or zip only if all the files are binary
   or the two systems use the same record format and character set for
   text files.

   Also seeSections 4.3,4.10,4.11, and4.15 for how
   to transfer directory trees between both like and unlike systems
   directly with Kermit.

4.2.7. Pipes and Encryption

   Of course pipelines could be used for encrypted file transfers,
   assuming proper precautions could be taken concerning the transmission
   of the key. But there is rarely a good way to do this. To illustrate
   using UNIX crypt:

  csend {crypt key < filename} {!crypt key > filename}

   Or, more ambitiously:

  csend {tar cf - . | gzip -c | crypt key} {!crypt key | gunzip -c | tar xf -}

   transmits the key in the file header packet as part of the (clear-text)
   remote command, defeating the entire purpose of encrypting the file
   data.

   But if you are connected in terminal mode to the remote computer and
   type:

  creceive {crypt key > filename}

   at the remote Kermit prompt, you have also transmitted the key in clear
   text over the communications link.

   At present, the only secure way to use CSEND and CRECEIVE with an
   encryption filter is to have a human operator at both ends, so the key
   does not have to be transmitted.

   Theoretically it would be possible to use PGP software (Pretty Good
   Privacy, by Phil Zimmerman, Phil's Pretty Good Software) to avoid key
   transmission (since PGP uses separate public and private key and "lets
   you communicate securely with people you've never met, with no secure
   channels needed for prior exchange of keys"), but the specific method
   has yet to be worked out.

     HINT: See the PGP User's Guide, e.g. at:
    http://www.telstra.com.au/docs/PGP/

     Especially the topic "Using PGP as a UNIX-Style Filter":
    http://www.telstra.com.au/docs/PGP/pgpdoc2/pgpdoc2_17.html

   In any case, better and more convenient security options are now
   available: Kerberos authentication and encryption and the new
   ability to run C-Kermit "though" other communication programs,
   described in Section 2.7.

4.2.8. Commands and Functions Related to Pipes

4.2.8.1. The OPEN !READ and OPEN !WRITE Commands

   These are described inUsing C-Kermit, and are generally useful
   with reading output from commands that produce more than one line on
   their standard output, or writing multiple lines into commands that
   accept them on their standard input.

   In C-Kermit 7.0 CLOSE !READ is accepted as a synonym for CLOSE READ,
   and CLOSE !WRITE for CLOSE WRITE.

   Testing the success and failure of these commands, however, can be a
   bit tricky. Consider:

  open !read lalaskjfsldkfjsldkfj

   (where "lalaskjfsldkfjsldkfj" is neither a valid command nor the name
   of a program or script that can be run). OPEN !READ, in UNIX at least,
   translates this into execl(shellpath,shellname,"-c",command). This
   means it starts your preferred shell (e.g. from the SHELL environment
   variable) and asks it to execute the given command. It must be this
   way, because your command can be a either an internal shell command
   (which only your shell can execute) or an external command, which only
   your shell knows how to find (it knows your PATH and interprets, etc).
   Therefore unless OPEN !READ can't start your shell, it always succeeds.

   Continuing with the nonexistent-command example:

  C-Kermit> open !read lalaskjfsldkfjsldkfj
  C-Kermit> status
   SUCCESS
  C-Kermit> read line
  C-Kermit> status
   SUCCESS
  C-Kermit> echo "\m(line)"
  "bash: lalaskjfsldkfjsldkfj: command not found"
  C-Kermit> close read
  C-Kermit> status
   FAILURE
  C-Kermit>

   In other words, the failure can not be detected on OPEN, since the OPEN
   command succeeds if it can start your shell. It can't be detected on
   READ, since all this does is read output from the shell, which in this
   case happens to be an error message. However, failure IS detected upon
   close, since this is the occasion upon which the shell gives Kermit its
   exit status code.

   For an illustration of this situation, see Section 2.14.

4.2.8.2. The REDIRECT Command

   A second method of I/O redirection is offered by the REDIRECT command.
   This is a rather advanced and tricky feature that is presently
   supported only in UNIX C-Kermit, in OS-9 C-Kermit, and in Kermit 95.
   Syntax:

   REDIRECT command
          Runs the given command, sending its standard output to the
          current communications channel (SET LINE, SET PORT, or SET HOST
          connection), and reading its standard input from the same
          connection. Works only in local mode -- i.e. a connection is
          required -- and then only if the given command uses Standard
          I/O.

   Example:

  redirect finger

   runs the local "finger" command and sends its output over the
   connection as plain text, where presumably there is a process set up to
   read it. Another example:

  redirect finger | sort -r

   shows the use of a pipeline.

   Note: REDIRECT differs from CSEND/CRECEIVE in two important ways: (1)
   it does not use the Kermit protocol, and (2) it uses a bidirectional
   pipe rather than a one-way pipe.

   The primary use of the REDIRECT command is to run external protocols,
   such as sz/rz in UNIX for ZMODEM, when they work over Standard I/O(*).
   Example:

  set host xyzcorp.com
  (login, etc)
  redirect sz oofa.zip

   lets you make a Telnet connection with C-Kermit and then do a ZMODEM
   transfer over it. ZMODEM protocol messages go both ways over the same
   connection simultaneously.

   It is possible to use C-Kermit on UNIX as your PPP dialer and then to
   REDIRECT the connection to the PPP software, but C-Kermit 7.0 offers a
   better approach to PPP dialing in its new EXEC command (Section
   1.23).

   In theory, you can also redirect an interactive process. For example,
   suppose you tell Kermit 95 to wait for an incoming TCP/IP connection:

  set host * 3000

   and then tell C-Kermit on UNIX to:

  set host kermit95hostname 3000
  redirect ksh

   and then tell Kermit 95 to CONNECT: now you are talking to the UNIX
   K-shell; you can give commands (pwd, ls, etc) and see the results. In
   practice, the K-shell's terminal modes are messed up because (a) it is
   not going through the Unix terminal driver, and (b) it is "smart" and
   knows it is being redirected, and so acts in a decidedly inhospitable
   manner (other applications like EMACS, vi, etc, simply refuse to run if
   their standard i/o has been redirected).

     (*) The publicly-distributed sz/rz programs do not work as clients.
     However, Omen Technology does offer an up-to-date redirectable
     client XYZMODEM program called crzsz.

4.2.8.3. Receiving Mail and Print Jobs

   As of 7.0, and in UNIX only, files that are sent to C-Kermit as mail
   (when the other Kermit uses a MAIL or SEND /MAIL command) or to be
   printed (via REMOTE PRINT or SEND /PRINT) are now piped directly to the
   mail or print program, rather than written to temporary files and then
   mailed or printed and then deleted. This has the advantages of (a) not
   requiring a temporary file, and (b) allowing mail to have a proper
   subject in place of the filename. Temporary files were bad not only
   because they required (a) space, and (b) writability of the current
   directory, but also because using them could result in wiping out an
   existing file. See Section 4.7 for more about SEND /MAIL and SEND
   /PRINT.

4.2.8.4. Pipe-Related Functions

   The \fcommand(command) function runs the given shell or system command
   and returns the command's standard output as its value (with any
   newline characters stripped from the end), unless the result is too
   long, in which case it returns the empty string. The maximum length for
   the result is at least 1022 bytes, and it might be longer on some
   platforms. Examples (UNIX):

  C-Kermit> echo "\fcommand(date)"
  "Fri Apr 18 13:31:42 1997"
  C-Kermit> echo "\fcommand(finger | wc -l)" ; how many users logged in?
  "      83"
  C-Kermit> evaluate \fcommand(finger | wc -l) * 2
  166
  C-Kermit> echo Welcome to \fcommand(tty) on \fcommand(date)
  Welcome to /dev/ttyre on Fri Apr 18 13:31:42 1997
  C-Kermit> echo "\fcommand(ls oofa.*)"
  "oofa.c
  oofa.h
  oofa.o"
  C-Kermit> cd /directory-with-thousands-of-files
  C-Kermit> echo "\fcommand(ls -l)" ; This would be too long
  ""
  C-Kermit>

   If a command's output would be too long, you can use the other, more
   laborious method of reading from a command: OPEN !READ command, READ
   each line, CLOSE !READ.

   The \frawcommand(command) function is identical to \fcommand(command),
   except it does not remove trailing newline characters:

  C-Kermit> echo "\frawcommand(date)"
  "Fri Apr 18 13:31:42 1997
  "
  C-Kermit> echo "\frawcommand(ls oofa.*)"
  "oofa.c
  oofa.h
  oofa.o
  "
  C-Kermit>

   Use \frawcommand() if you want to retain the final line terminators, or
   if the command's output is "binary". But remember that if the result of
   this (or any other) function contains any NUL (ASCII code 0)
   characters, the first NUL will terminate the result string because this
   is how C strings work (it's "C-Kermit", remember?).

   These functions are useful not only locally, but also in the
   client/server arena. If you need to get the results from a system
   command on the server end into a variable on the client end, just do:

  [ remote ] query kermit command(date)

   The result is in the local \v(query) variable; seeUsing C-Kermit,
   2nd Ed., pp.359-360 for details.

4.3. Automatic Per-File Text/Binary Mode Switching

   When transferring files between like systems (e.g. UNIX-to-UNIX),
   binary mode can be used for all files unless character-set translation
   is needed, and in fact Kermit programs of recent vintage recognize each
   others' platforms and switch to binary mode automatically when it is
   appropriate (e.g. DOS to OS/2, or UNIX to UNIX). (Exception: LABELED
   mode is chosen for VMS-to-VMS and OS/2-to-OS/2 transfers so complex
   file formats can be preserved.)

   On a client/server connection between like systems, the transfer mode
   is currently determined by the file sender, rather than always by the
   client. If the client is sending, it controls the transfer mode. If a
   GET command is sent to the server, the server sends all files in binary
   mode if its TRANSFER CHARACTER-SET is TRANSPARENT; otherwise it uses
   text mode for text files (according to its text-pattern list) and
   binary mode for binary files. Of course, the client can control the
   server's transfer character-set with the REMOTE SET TRANSFER
   CHARACTER-SET command.

   When transferring files between unlike systems, however, (e.g.
   UNIX-to-DOS), some files (such as executable program images) must be
   transferred in binary mode but others (such as plain-text files) must
   be transferred in text mode so their record format and character sets
   can be appropriately converted. If a binary file is transferred in text
   mode, it is ruined. If a text file is transferred in binary mode, then
   at the very least, its format can be incorrect; at worst it is also
   corrupted because its character set was not converted (in extreme cases
   the corruption is total, e.g. because one system is ASCII-based and the
   other EBCDIC).

4.3.1. Exceptions

   VMS C-Kermit, when sending files to a non-VMS system, switches to text
   or binary mode automatically for each file, based on the record format
   in the file's directory entry; thus the mechanisms described in this
   section do not apply to VMS C-Kermit, yet the effect is the same:
   automatic text/binary mode switching when VMS C-Kermit is sending
   files. See the VMS Appendix ofUsing C-Kermit for details.

   Kermit versions that support LABELED or IMAGE transfer mode are
   likewise not affected by this feature when one of those modes is
   selected (normally used only when transferring between like systems).

   Kermit versions that support file-transfer pipes and filters are not
   affected by this feature when pipes or filters are used, since the
   output of a pipe or filter (such as gzip) is likely to require transfer
   in a different mode than the original file.

   Finally, SEND /TEXT or SEND /BINARY will force files to be sent in the
   indicated mode, overriding all automatic transfer-mode-choosing
   mechanisms.

4.3.2. Overview

   Suppose you give C-Kermit a command like:

  SEND *.*

   And suppose the pattern *.* matches a mixture of text files (such as
   program source code) and binary files (such os object modules or
   executable programs).

   C-Kermit 6.0 and earlier (except on VMS) send all files in the same
   mode: whatever you said in your most recent SET FILE TYPE command, or
   else whatever mode was chosen automatically according to the rules on
   page 236 of Using C-Kermit, 2nd Ed.

   But when text and binary files are mixed in the same group, and the
   files are being transferred to an unlike system (e.g. UNIX to IBM
   Mainframe), this results in corruption of either all the text files or
   all the binary files.

   Stream-oriented file systems such as in UNIX and DOS do not record any
   information about the file to tell us whether the file should be
   transferred in binary or text mode, making it impossible to select the
   transfer mode for each file in a group automatically with any
   certainty.

   However, we can use some fairly-well established file naming
   conventions for this purpose. C-Kermit 7.0 lets you provide lists of
   filename patterns that are used to separately determine the file type
   for each individual file being transferred. A pattern is a string,
   possibly containing the special characters "*" (asterisk, which matches
   any string of zero of more characters) and/or "?" (question mark, which
   matches any single character). For example "a*b" matches all files
   whose names start with "a" and end with "b", such as "ab", "arb",
   "ababababab", etc, but not "abba". And "a?b" matches any file whose
   name starts with "a", ends with "b", and is exactly 3 characters long.

     NOTE: When typing commands at the C-Kermit prompt, you must prefix
     "?" with \ to override its normal function of giving help.

   (Also see Section 4.9 for additional pattern-matching notations
   that might be available in your version of C-Kermit.)

   When you have specified filename recognition patterns, C-Kermit can
   transfer the ones whose names match any of the binary-mode patterns in
   binary mode, and those with names that match any of the text-mode
   patterns in text mode, and those whose names match neither in the
   prevailing mode you have chosen, or that was chosen automatically via
   peer recognition.

4.3.3. Commands

   SET FILE PATTERNS { ON, OFF, AUTO }
          This tells Kermit whether to do per-file filename
          pattern-matching to determine text or binary mode. The normal
          and default setting is AUTO, which means to use pattern lists to
          switch transfer mode only when it is certain that the other
          Kermit program supports automatic notification of transfer mode
          (via Attribute packets) on a per-file basis (this information is
          obtained automatically during protocol startup negotiation). ON
          means to always determine the transfer mode from the filename
          and pattern list when sending files. Use OFF to disable this
          feature (without resetting your pattern lists). Also note that
          if you have selected LABELED file transfer (SET FILE TYPE
          LABELED), this takes precedence over filename-matching patterns
          and all files are sent in labeled mode.

   SET TRANSFER MODE MANUAL
          Disables the use of filename patterns, no matter what the FILE
          PATTERNS setting.

   REMOTE SET TRANSFER MODE MANUAL
          Client command to disable automatic transfer mode, and therefore
          also filename patterns, in the server. Synonym: REMOTE SET XFER
          MODE MANUAL.

   { GET, SEND, etc } { /BINARY, /TEXT }
          Including a /BINARY or /TEXT (or, where supported, /IMAGE or
          /LABELED) switch with a file-transfer command changes the
          transfer mode to manual for that command only, and therefore
          disables patterns that that command.

   SET FILE BINARY-PATTERNS [ pattern [ pattern [ pattern ... ] ] ]
          A list of zero or more patterns, separated by spaces (not
          commas). Letters in a pattern are case-sensitive if the
          underlying filenames are case sensitive (as in UNIX), and
          case-insensitive otherwise (as in Windows). If a file's name is
          matched by any pattern in the list and SET FILE PATTERNS is ON,
          the file is sent in binary mode. Examples:

  SET FILE BINARY-PATTERNS *.gz *.Z *.tar *.zip *.o *.so *.a *.out ; UNIX
  SET FILE BINARY-PATTERNS *.EXE *.ZIP *.OBJ *.COM ; DOS or OS/2 or Windows

          If a pattern contains spaces, enclose it in braces.

   SET FILE TEXT-PATTERNS [ pattern [ pattern [ pattern ... ] ] ]
          Like SET FILE BINARY-PATTERNS, but the patterns choose text
          files rather than binary ones. Examples:

  SET FILE TEXT-PATTERNS *.TXT *.KSC *.HTM* *.BAT ; DOS, Windows, OS/2

   ADD BINARY-PATTERNS [ pattern [ pattern [ pattern ... ] ] ]
          Adds one or more patterns to the BINARY-PATTERN list.

   ADD TEXT-PATTERNS [ pattern [ pattern [ pattern ... ] ] ]
          Adds one or more patterns to the TEXT-PATTERN list.

   REMOVE BINARY-PATTERNS [ pattern [ pattern [ pattern ... ] ] ]
          Removes one or more patterns from the BINARY-PATTERN list. The
          given patterns are matched with the ones in the BINARY-PATTERNS
          list with case sensitivity if the underlying file system has
          case-sensitive names (as do UNIX and OS-9), otherwise with case
          independence.

   REMOVE TEXT-PATTERNS [ pattern [ pattern [ pattern ... ] ] ]
          Removes one or more patterns from the TEXT-PATTERN list.

   SHOW PATTERNS
          Displays the current pattern selections.

   Whenever you give a SET FILE BINARY-PATTERNS or SET FILE TEXT-PATTERNS
   command, the previous list is replaced. If you give one of these
   commands without a pattern list, the previous list is removed.

   When patterns are active and files are being sent, text patterns (if
   any) are applied first (but only if not RESENDing and not sending in
   LABELED mode), then binary patterns, so if the same pattern appears in
   both lists, binary mode is chosen.

4.3.4. Examples

   Here's an example that might be used when sending files from UNIX:

  set file type binary
  set file text-patterns *.c *.h *.w *.txt makefile
  set file binary-patterns *.o
  msend makefile wermit wart ck*.[cwho] ck*.txt

   Note that "wermit" and "wart" do not match any patterns so they are
   sent in the prevailing mode, which is binary. Also note the use of
   "makefile" as a pattern that does not contain any wildcard characters
   (there is no other convention to distinguish among "wermit" and "wart",
   which are binary executables, and "makefile", which is a text file,
   purely by their names).

   Most C-Kermit implementations have a default pattern list built in,
   which includes patterns that are almost certain to succeed in picking
   the right transfer mode. Others are omitted due to ambiguity. For
   example ".hlp", and ".ini" are generally binary types in Windows but
   text types everywhere else.

     NOTE: ".doc", used for decades to denote plain-text documentation
     files, now more often than not denotes a Microsoft Word Document, so
     ".doc" is now considered a binary type since it does less harm to
     transfer a plain-text document in binary mode than it does to
     transfer an MS Word file in text mode (except when IBM mainframes
     are involved!)

     ANOTHER NOTE: ".com" files are binary in DOS-like operating systems,
     but they are text (DCL command procedures) in VMS. VMS C-Kermit
     sends .COM files in text mode; K95 sends them in binary mode. If you
     download a .COM file from VMS to DOS or Windows, and then upload it
     to another VMS system, be sure to use SEND /TEXT to preserve its
     textness.

   You can see the default pattern list by starting C-Kermit without its
   initialization file (e.g. "kermit -Y") and using the SHOW PATTERNS
   command. If you will be depending on this feature, be sure to examine
   the list carefully in conjunction with the applications that you use.

   The default pattern list does not take "backup files" into account
   because (a) people usually don't want to transfer them; and (b) it
   would make the pattern lists more than twice as long. For example, we
   would need to include both *.txt and *.txt.~[0-9]*~ for ".txt" files,
   and similarly for all the others. Instead, you can use SEND /NOBACKUP
   (or SET SEND BACKUP OFF) to skip over all backup files.

   Put your most commonly-used safe pattern declarations in your C-Kermit
   customization file (ckermod.ini, .mykermrc, k95custom.ini, etc).

   As noted, SET FILE PATTERNS is ON by default. Sometimes, however, it is
   desirable, or necessary, to force files to be sent in a particular
   mode, and often this must be done from the command line (e.g. when
   using Kermit as a download helper in a Web browser like Lynx). The -V
   command-line options is equivalent to SET FILE PATTERNS OFF and SET
   TRANSFER MODE MANUAL. Example:

  kermit -Vis oofa.txt

   forces oofa.txt to be sent in binary mode, even though ".txt" might
   match a text pattern.

4.4. File Permissions

   "Permissions" refers to a code associated with a file that specifies
   who is allowed to access it, and in what manner. For example, the
   owner, the members of one or more groups, the system administrator, and
   everybody else, might be allowed various combinations of Read, Write,
   Append, Execute, or Listing access.

   The permission code goes by different names on different platforms. In
   UNIX, it might be called the filemode. In VMS, it is called the file
   protection (or protection mask).

   The comments in this section presently apply only to the UNIX and VMS
   versions of C-Kermit, to which these features were added in version
   7.0; the DOS, Windows, and OS/2 file systems embody no notions of
   protection, and so MS-DOS Kermit and Kermit 95 do not send file
   permissions, and ignore them when received.

   The permissions for a received file are determined by a combination of
   the file transfer mode (VMS-to-VMS transfers only), whether a file of
   the same name exists already, whether permissions of the file are
   received in the file attribute packet, and the setting of ATTRIBUTES
   PROTECTION.

   The default for ATTRIBUTES PROTECTION is ON. If no attributes are
   received, the effect is the same as if attributes PROTECTION were OFF.

   For VMS-to-VMS transfers, the default LABELED mode simply copies the
   protection code from source to destination.

4.4.1. When ATTRIBUTES PROTECTION is OFF

   If no file of the same name exists, system defaults determine the
   permissions of the new file. Otherwise, the actions taken depend on the
   current FILE COLLISION setting: BACKUP, OVERWRITE, RENAME, etc, as
   documented inUsing C-Kermit. But now the new file (if it is
   created at all) automatically inherits the permissions (mode bits) of
   the existing file in a way that is appropriate for the platform.

4.4.1.1. Unix

   All mode bits are inherited except the directory bit, since the
   incoming file can not possibly be a directory. (In any case, it is not
   possible to receive a file that has the same name as an existing
   directory unless FILE COLLISION is set to RENAME).

4.4.1.2. VMS

   Files with the same name as an existing file, transferred in modes
   other than LABELED between VMS systems, inherit the protection of the
   prior version.

4.4.2 When ATTRIBUTES PROTECTION is ON

   File permissions can be conveyed as part of the file transfer process,
   in accordance with the Kermit protocol definition. If the file sender
   puts system-dependent and/or system-independent versions of the file
   protection (permissions) into the Attribute (A) packet, the file
   receiver can set the new file's permissions from them. Otherwise, the
   permissions are set the same as for ATTRIBUTES PROTECTION OFF.

   When the incoming A packet contains system-dependent permissions, the
   file receiver checks to see if the sender has the same system ID (e.g.
   both the sending and receiving systems are UNIX, or both are VMS); if
   so, it decodes and uses the system-dependent permissions; otherwise it
   uses the generic ones (if any) and applies them to the owner field,
   setting the other fields appropriately as described in the following
   sections.

   Setting the incoming file's protection from the A packet is controlled
   by SET ATTRIBUTES PROTECTION (or PERMISSION), which is ON by default,
   and its status is displayed by SHOW ATTRIBUTES.

   The main benefit of this feature is to not have to "chmod +x" an
   executable file after transfer from UNIX to UNIX. Its cross-platform
   benefits are less evident, perhaps to retain the status of the Unix 'x'
   bit on a VMS system, for subsequent transfer back to a Unix system.

4.4.2.1. System-Specific Permissions

   System-specific file permissions are used when the two Kermit programs
   recognize each other as running on the same type of system. For
   example, both are running under some form of UNIX (it doesn't matter
   which UNIX variation -- HP-UX, Solaris, AIX, etc -- all use the same
   scheme for file permissions); or both are running under VMS (even if
   one is on an Alpha and the other on a VAX, and/or one is old and the
   other is new).

4.4.2.1.1. UNIX

   UNIX supports three categories of users, File Owner, Group, and World,
   and three types of file access permission: Read, Write, and Execute.
   Thus, a UNIX file's permissions are expressed in 9 bits.

   The system-dependent permission string for UNIX is a 3-digit octal
   string, the low-order 9 bits of the st_mode member of the stat struct;
   we deliberately chop off the "file format" bits because they are not
   permissions, nor do we convey the setuid/setgid bits, lock bit, sticky
   bit, etc.

4.4.2.1.2. VMS

   VMS supports four categories of users, System, File Owner, Group, and
   World, and four types of file access permission: Read, Write, Execute,
   and Delete. Thus, a VMS file's permissions are expressed in 16 bits.

   The system-dependent protection string for VMS is a 4-digit hexadecimal
   string, corresponding to the internal-format protection word of the
   file (RWED for each of World,Group,Owner,System). A new file normally
   gets all 16 protection bits from the original file of the same name.

   Note: VMS-to-VMS transfers take place in LABELED mode when the two
   C-Kermits recognize each other's platform as VMS (unless you have
   disabled LABELED-mode transfers). In this case, all of a file's
   attributes are preserved in the transfer and the protection mask (and
   other information) is taken from the file's internal information, and
   this takes precedence over any information in the Attribute packets.
   You can defeat the automatic switching into LABELED mode (if you want
   to) with SET TRANSFER MODE MANUAL.

4.4.2.2. System-Independent Permissions

   The system-independent ("generic") protection is used when the system
   IDs of the two Kermit programs do not agree (e.g. one is UNIX, the
   other is VMS). The generic protection attribute includes the following
   permissions (not all are applicable to every file system): Read, Write,
   Append, Execute, Delete, Search. The generic permissions are derived
   from the owner permissions of the source file, thus, a Unix 'w'
   permission becomes VMS Write,Delete.

   The Owner field of the new file's permissions is set from the incoming
   generic protection attribute.

   In UNIX, the Group and World permissions are set according to your
   umask, except that execute permission is NOT set in these fields if it
   was not also set in the generic protection (and consequently, is set in
   the Owner field).

   In VMS, the System, Group, and World permissions are set according to
   the process default file permission (as shown in VMS by SHOW
   PROTECTION), except that no permissions are allowed in these fields
   that are not included in the generic permissions.

   Note that the VMS and UNIX interpretations of Execute permission are
   not identical. In UNIX, a file (binary executable, shell script, etc)
   may not be executed unless it has Execute permission, and normally
   files that are not intended for execution do not have Execute
   permission. In VMS, Read permission implicitly supplies Execute
   capability. Generally files that have Read permission also have
   explicit Execute permission, but files (binary executables, DCL command
   procedures) that have Read permission and not Execute permission can
   still be executed.

4.5. File Management Commands

4.5.1. The DIRECTORY Command

   Prior to C-Kermit 7.0, the DIRECTORY command always ran an external
   system command (such as "ls" on UNIX) or program to product the
   directory listing. This had certain advantages, mostly that you could
   include system-dependent options for customized listings, e.g. on UNIX:

  dir -lt c* | more

   or in VMS:

  directory /size/date/protection/except=*.obj oofa.*;0

   This approach, however, carries some disadvantages: C-Kermit can't
   return SUCCESS or FAILURE status for (e.g.) "dir foo" according to
   whether the file "foo" exists; and it runs an inferior process, which
   might be a problem in some environments for resource and/or security
   reasons, and won't work at all in a "nopush" environment (e.g. one in
   which C-Kermit is configured to forbid access to exterior commands and
   programs, e.g. in a VMS "captive account").

   In C-Kermit 7.0 on VMS and UNIX, and in K95 1.1.19 and later, the
   DIRECTORY command is internal to Kermit. It can be run in a "nopush"
   environment and returns SUCCESS or FAILURE status appropriately. In
   UNIX it prints all dates and times in a consistent way (unlike ls). In
   VMS it prints precise file sizes, rather than "blocks". It offers
   several formatting and other options, but it is not necessarily more
   flexible than the corresponding external commands or programs (the UNIX
   "ls" program, the VMS "directory" command). The syntax is:

   DIRECTORY [ switch [ switch [ ... ] ] ] [ filespec ]

   If no filespec is given, all files in the current directory are listed.

   Optional switches include all the standard file-selection switches
   presented in Section 1.5.4, plus:

   /ALL
          Show both regular files and directories; this is the default.

   /ARRAY:x
          Instead of displaying a directory listing, put the files that
          would have been shown (based on the filespec and other selection
          switches) in the given array. The array need not (and should
          not) be predeclared; if the array already exists, it is
          destroyed and reused. The array name can be a single letter,
          like "a", or any fuller form, such as "&a", "\&a", "\&a[]", etc.
          If the /ARRAY switch is included, the following other switches
          are ignored: /BRIEF, /VERBOSE, /HEADING, /PAGE, /ENGLISHDATE,
          /ISODATE, /XFERMODE, /MESSAGE, /SORT, /REVERSE, /ASCENDING. In
          other words, only file selection switches are meaningful with
          /ARRAY: /FILES, /DIRECTORIES, /ALL, /DOTFILES, /NOBACKUP,
          /RECURSIVE, /SMALLER, /LARGER, /AFTER, /BEFORE, /EXCEPT, etc.
          The resulting array has the number of files (n) as its 0th
          element, and the filenames in elements 1 through n Example:

  dir /array:&a /files /nobackup /after:19990101 /larger:10000 [ab]*
  show array &a

   /FILES
          Only show regular files.

   /DIRECTORIES
          Only show directories.

   /BACKUP
          In UNIX, OS-9, K-95, and other versions that support SET FILE
          COLLISION BACKUP and create backup files by appending .~n~ to
          the filename (where "n" is a number), /BACKUP means to include
          these files in directory listings. This is the default.

   /NOBACKUP
          This is the opposite of /BACKUP: that is, do not include backup
          files in the listing.

   /BRIEF
          List filenames only; use a compact format, as many filenames as
          will fit across the screen (based on the longest name). A brief
          listing is always sorted alphabetically.

   /VERBOSE
          List one file per line, and include date, size, and (in UNIX
          only) permissions of each file. This is the opposite of /BRIEF,
          and is the default.

   /PAGE
          Pause at the end of each screenful and give a "more?" prompt,
          even if SET COMMAND MORE-PROMPTING is OFF.

   /NOPAGE
          Don't pause at the end of each screenful and give a "more?"
          prompt, even if SET COMMAND MORE-PROMPTING is ON. If neither
          /PAGE or /NOPAGE is given, paging is according to the prevailing
          COMMAND MORE-PROMPTING setting (which can be displayed with SHOW
          COMMAND).

   /ENGLISHDATE
          Show dates in dd-mmm-yyyy format; mmm is the first three letters
          of the English month name.

   /ISODATE
          Show dates in yyyy-mm-dd format; mm is the month number, 1-12.
          This is the opposite of /ENGLISHDATE, and is the default.

   /HEADINGS
          Print a heading before the listing and a summary at the end.

   /NOHEADINGS
          Don't print a heading before the listing or a summary at the
          end. This is the opposite of /HEADINGS, and is the default.

   /XFERMODE
          Only in Kermit programs that support SET FILE PATTERNS. If this
          switch is included, and the filename matches any FILE
          BINARY-PATTERN ( Section 4.3), "(B)" is printed after the
          filename; otherwise, if it matches a FILE TEXT-PATTERN, "(T)" is
          printed.

   /NOXFERMODE
          Don't display transfer-mode indicators. This is the opposite of
          /XFERMODE and is the default.

   /RECURSIVE
          Show files not only in the given directory, but also in its
          subdirectories (if any), their subdirectories, etc.

   /NORECURSIVE
          Don't show files in subdirectories. This is the opposite of
          /RECURSIVE, and is the default.

   /MESSAGE:text
          This lets you specify a short text string to be appended to the
          end of each directory listing line (a space is supplied
          automatically). If the text contains any spaces, enclose it in
          braces, e.g. /MESSAGE:{two words}.

   /NOMESSAGE
          Don't append any message to the end of each directory listing
          line (default).

   /SORT:[{NAME,SIZE,DATE}]
          Sort the listing by name, size, or date. If the /SORT switch is
          given but the "sort-by" keyword is omitted, the listing is
          sorted by name. /SORT:NAME /ASCENDING (alphabetic sort by name)
          is the default.

   /NOSORT
          Don't sort the listing. Files are listed in whatever order they
          are supplied by the operating system, e.g. inode order in UNIX.

   /REVERSE
          If the /SORT switch is given, reverse the order of the sort.
          Synonym: /DESCENDING.

   /ASCENDING
          If the /SORT switch is given, sort the listing in normal order.
          This is the opposite of /REVERSE and is the default.

   Note that most of the DIRECTORY-specific switches come in pairs, in
   which one member of a pair (e.g. /NOHEADINGS) is the opposite of the
   other (e.g. /HEADINGS).

   If you always want to use certain options, you can set them with the
   SET OPTIONS DIRECTORY command ( Section 1.5.5). Use SHOW OPTIONS to
   list the options currently in effect. To make the desired options apply
   every time you run C-Kermit, put a SET OPTIONS DIRECTORY command in
   your C-Kermit customization file, specifying the desired options.
   Options set in this manner apply to every subsequent DIRECTORY command.
   Of course, if you include switches in a DIRECTORY command, these
   override any defaults, built-in or custom. Example:

  DIRECTORY            ; Use "factory defaults"
  SET OPTIONS DIRECTORY /SORT:SIZE /REVERSE /HEADINGS  ; Customize defaults
  DIRECTORY            ; Use customized defaults
  DIR /SORT:NAME       ; Override customized default SORT key
  SET OPT DIR /RECURS  ; Add /RECURSIVE to customized defaults
  DIR /ASCEND          ; Override customized default SORT order

   Notes:

     * Only a single sort key is supported; there is presently no way to
       have multiple sort keys.
     * If the /BRIEF switch is given, all other switches (except
       /[NO]RECURSIVE, /[NO]DOTFILES, /DIRECTORIES, /FILES, and /ALL) are
       ignored.
     * /SORT:anything gives incorrect results if any files have lengths
       greater than 10 digits (i.e. that are more than 9999999999 bytes
       long, i.e. if they are 10GB or more in size) because the overlong
       length field causes the date and name fields to be misaligned.
     * /SORT:NAME is redundant in VMS since VMS returns filenames in
       alphabetic order anyway.
     * /SORT:NAME ignores alphabetic case on platforms where case does not
       matter in filenames, but this works only for unaccented Roman
       letters A-Z.
     * /SORT:NAME is currently based on code values, and so works fine for
       ASCII, but will probably produce unexpected results for files with
       non-ASCII or 8-bit characters in their names. (Locale-based sorting
       raises rather significant issues of portability, size, performance,
       etc.)
     * /SORT:DATE works right only for ISO-format dates, not English ones.
     * /SORT:SIZE sorts the size field lexically. On some platforms (e.g.
       Windows), the size of a directory file is listed as "<DIR>" rather
       than as a number; in this case, the "<DIR>" files are gathered at
       the end (or beginning, depending on the sort order) of the listing.
     * /RECURSIVE is accepted but ignored in AOS/VS. Use the normal
       system-specific filespec notation, e.g. "dir #.txt".
     * /RECURSIVE has no affect when a full, absolute pathname is given;
       e.g. "dir /recursive /tmp/foo" (where "foo" is a regular file) only
       shows the "/tmp/foo" file. If you want to see all "foo" files in
       the /tmp tree, do "cd /tmp" and then "dir /recursive foo".
     * If a file size of -1 is shown, or date-time of 0000-00-00 00:00:00,
       this means the file was located, but access to information about
       the file was denied to C-Kermit.
     * In VMS, if FOO.DIR;1 is a directory within your current directory,
       "directory foo" and "directory [.foo]" list the files in the [.FOO]
       subdirectory, but "directory foo.dir" lists the directory file
       itself; similarly for "*.dir" versus "[.*]", etc.
     * In UNIX, if "foo" is a directory within your current directory,
       "directory foo" lists the files in the foo directory. If you want
       to list the foo directory file itself, put an asterisk at the end:
       "dir foo*".

   Hint: How to find the biggest files in a directory tree:

  cd xxx ; (root of tree)
  directory /sort:size /recursive /reverse /dotfiles /page

   Another hint: If you often use several different directory-listing
   formats, define macro shortcuts for them:

  DEFINE WD DIRECTORY /SORT:DATE /REVERSE \%*  ; Reverse chronological order
  DEFINE SD DIRECTORY /SORT:SIZE /REVERSE \%*  ; Reverse order of size
  DEFINE ND DIRECTORY /SORT:NAME /ASCEND \%*   ; Alphabetical by name
  DEFINE DL DIR /DIR /SORT:NAME /ASCEND \%*    ; Alphabetical directory list

   Put these definitions in your C-Kermit customization file. Note that
   "\%*" ( Section 7.5) in these definitions lets you include other
   switches in your macro invocations, e.g.:

  wd /headings *.txt

   Of course you can still access your external directory listing program
   by using RUN or "!", e.g. in VMS:

  run directory /size/date/protection/except=*.obj oofa.*;0

   or:

  !dir /size/date/prot/exc=*.obj oofa.*;0

   In UNIX, use "!ls" or just "ls" (which is a special synonym for "!ls").

4.5.2. The CD and BACK Commands

   In C-Kermit 7.0, the CD command has a new friend, the BACK command.
   BACK means "CD to my previous current directory". A second BACK brings
   you back to where you were before the first one; thus successive BACK
   commands switch back and forth between two directories.

4.5.2.1. Parsing Improvements

   The CD command, as well as other commands that parse a directory name,
   were changed in 7.0 to provide all the expected functions: completion
   on Tab or Esc, directory-name lists on ?, etc. Other affected commands
   include SET SERVER GET-PATH, SET TEMP-DIRECTORY, SET FILE
   DOWNLOAD-DIRECTORY, and SPACE. CD and REMOTE CD also now work with
   logical names.

   In VMS, the situation is a bit complicated since a directory name can
   look like "DEV:", "[FOO.BAR]", "DEV:[FOO.BAR]", "[FOO]BAR.DIR;1", etc.
   Completion and ?-help might not always work, but they do in many cases.
   Examples:

  cd ?           Lists all subdirectories of the current directory
  cd []?         Ditto
  cd k?          Ditto, but only those starting with K
  cd [foo]?      Lists all subdirectories of the [FOO] directory
  cd [-]?        Lists all subdirectories of the superior directory
  cd [--]?       Lists all subdirectories of the directory 2 levels up
  cd [...]?      Lists all directories below the current one
  cd [foo.?      Does not work.

   C-Kermit allows all of the following in VMS:

  cd bar         CD to subdirectory BAR of the current directory
  cd .bar        Ditto
  cd [.bar]      Ditto
  cd bar.dir     etc...
  cd bar.dir;
  cd bar.dir;1
  cd [foo.bar]
  cd <foo.bar>
  cd bar.baz     This can go more than 1 level deep...
  cd dir:        (where logical name DIR is defined as [FOO.BAR])

   As well as the following:

  cd ..          Go up one level as in UNIX
  cd .           The current directory
  cd             My login directory

   Note that "cd -" (go up one level) does not work as expected, because
   "-" is Kermit's command continuation character. However, "cd [-]", and
   "
   cd {-}" have the desired effect (and so does "cd ..", which is easier
   to type).

4.5.2.2. The CDPATH

   The CD command in the UNIX, Windows, OS/2, and VMS versions of
   C-Kermit, as of version 6.1 / 1.1.12, searches the CDPATH for the given
   directory, if it is not absolute and if a CDPATH environment variable
   is defined. Example (in UNIX ksh or bash):

  $ export CDPATH=$HOME:$HOME/kermit:/tmp

   Now if you give a "cd xxx" command, no matter what your current
   directory is, if the "xxx" directory is not a subdirectory of your
   current directory, then the xxx subdirectory of your home directory is
   used or if that does not exist, then the xxx subdirectory of the kermit
   subdirectory of your home directory is used or if that does not exist,
   then /tmp/xxx is used. This is how the ksh "cd" command works, and now
   the C-Kermit CD command works the same way.

   In VMS, you can define CDPATH to be a list of directories that contain
   actual directory delimiters, and/or logical names representing
   directories, using commas to separate them, e.g.:

  $ define cdpath [HOME],[SOMEOTHERDIR],[HOME.MISC]
  $ define cdpath SYS$LOGIN:,DISK1:[HOME],DISK2:[SCRATCH.IVAN]

   Example:

  $ define cdpath SYS$LOGIN:,[IVAN],[OLAF],[OLGA.MISC]
  $ kermit
  DISK1:[OLGA] C-Kermit> cd blah

   tries the BLAH subdirectory of the user's login directory, then
   [OLGA.BLAH], [IVAN.BLAH], [OLAF.BLAH], and [OLGA.MISC.BLAH], in that
   order, using the first one it finds, failing if it finds none.

   In C-Kermit 7.0, you may also set the CDPATH from the Kermit prompt:

   SET CD PATH path
          Allows the CD PATH to be set from within C-Kermit.

   SHOW CD shows the CD path and all other information relevant to the CD
   command.

4.5.2.3. CD Messages

   Whenever you change directory, you can have C-Kermit display a "Read
   Me" file from the new directory automatically. The commands are:

   SET CD MESSAGE { ON, OFF, FILE list }
          ON enables this feature; OFF (the default) disables it. File
          lets you specify the name of the "Read Me" file. A list of names
          to look for can be given in the following format:

  {{name1}{name2}{name3}{...}}

          e.g.:

  SET SERVER CD-MESSAGE FILE {{./.readme}{README.TXT}{READ.ME}}

          The default list of CD-message files is system dependent.

   SHOW CD shows your current directory, previous directory, CD path, and
   CD message info.

4.5.3. Creating and Removing Directories

   The MKDIR command now allows you to create multiple directories at
   once:

  C-Kermit> mkdir a/b/c/d

   creates the directory a in the current directory (if it doesn't exist
   already), and then creates subdirectory b in the a directory (if it
   didn't exist already), and so on.

   If you use MKDIR to try to create a directory that already exists,
   C-Kermit will print a warning ("?Directory already exists"), but the
   MKDIR command will still succeed. If you want to avoid the warning
   message, use IF DIRECTORY first to check if the directory already
   exists.

   The RMDIR command, however, will not remove more than one directory,
   nor will it remove a directory that contains any files. (There is, as
   yet, no RMDIR /RECURSIVE command, although one might be added later.)

   In VMS, these commands (like CD) are more forgiving of your syntax than
   is the DCL command shell; "mkdir oofa" is equivalent to "mkdir [.oofa]"
   and so on. Also in VMS, you'll find that C-Kermit's RMDIR command is
   easier than deleting a directory in DCL, since it automatically first
   gives it owner delete permission if you are the owner.

4.5.4. The DELETE and PURGE Commands

   The DELETE command now offers a selection of switches, and has a new
   companion, the PURGE command. First, DELETE:

   DELETE [ switches... ] filespec
          Deletes the file or files that match the filespec, which may
          contain wildcards ( Section 4.9).

   Optional switches include the standard file-selection switches
   presented in Section 1.5.4, plus:

   /ASK
          Before deleting each file, ask permission interactively. Answers
          are Yes or OK (delete the file), No (don't delete it), or Quit
          (stop executing the DELETE command).

   /NOASK
          Don't ask permission to delete each file.

   /LIST
          List each file and show whether it was deleted. Synonyms: /LOG,
          /VERBOSE.

   /NOLIST
          Don't list files while deleting them. Synonyms: /NOLOG, /QUIET.

   /HEADING
          Print a heading and summary line.

   /NOHEADING
          Don't print a heading and summary line.

   /PAGE
          When listing, pause at the end of each screenful and give the
          "More?" prompt. If you reply "n" (no), the DELETE command
          terminates.

   /SIMULATE
          Do everything implied by the given switches and filespec, except
          do not actually delete any files. This lets you preview which
          files would be deleted; implies /LIST.

   Now the PURGE command:

   PURGE [ switches... ] [ filespec ]
          (VMS only) Runs the DCL PURGE command. Switches and filespec, if
          any, are passed directly to DCL without parsing or verification.
          Deletes excess versions of the given (or all) files. The rest of
          this section does not apply to VMS.

   PURGE [ switches... ] [ filespec ]
          (UNIX only) Deletes "backup files" that match the filespec,
          which may contain wildcards ( Section 4.9). If no filespec
          is given, all backup files in the current directory are selected
          (subject to modification by any switches). Do not include backup
          notation in the filespec.

   Explanation:

   To avoid destroying preexisting files when a new file arrives that has
   the same name, C-Kermit backs up the old file by appending a "backup
   number" to its name. In UNIX, the backup suffix consists of a period, a
   tilde, a number, and another tilde. For example, if a file called
   oofa.txt exists and a new oofa.txt file arrives, the original is
   renamed to oofa.txt.~1~. If another oofa.txt file arrives, the existing
   one is renamed to oofa.txt.~2~. And so on. This system is compatible
   with the one used by EMACS. Thus over time, if you receive a lot of
   files with C-Kermit or edit them with EMACS, backup files can build up.
   The new PURGE command lets you clean out accumulated backup files:

   Optional switches include the standard file-selection switches
   presented in Section 1.5.4, plus all the switches listed above for
   the DELETE command, plus:

   /KEEP:n
          Retains the n most recent (highest-numbered) backup files for
          each file. For example, if oofa.txt, oofa.txt.~1~, oofa.txt.~2~,
          oofa.txt.~10~, oofa.txt.~12~, and oofa.txt.~100~ exist, "purge
          /keep:2 oofa.txt" deletes oofa.txt.~1~, oofa.txt.~2~, and
          oofa.txt.~10~, and keeps oofa.txt, oofa.txt.~12~, and
          oofa.txt.~100~. If /KEEP is given without a number, one (the
          highest numbered) backup file is kept.

   CAUTION: The PURGE command should be used only when *.~*~ files truly
   are backup files. This is the case for EMACS, and it is the DEFAULT for
   C-Kermit. However, if C-Kermit's FILE COLLISION has been set to RENAME,
   newly received files will look like backup files. In that case, don't
   use the PURGE command or you'll be removing new files rather than old
   ones. (Use SHOW FILE to find the FILE COLLISION setting.)

   The PURGE command is presently available only in UNIX. The command
   succeeds if it deleted any files, or if it deleted no files but there
   were no errors. It fails if it deleted no files and there were errors
   (i.e. deletion was attempted but failed). In VMS, backup file versions
   are handled automatically by the OS, and a PURGE command can be used at
   the VMS prompt to clean them up.

   If you want certain switches to be supplied automatically with each
   DELETE or PURGE command, you can set them with SET OPTIONS
   ( Section 1.5.5) and you can display any such settings with SHOW
   OPTIONS. Of course you can override them on a per-command basis by
   including switches in your PURGE or DELETE command.

   Also see SET FILE COLLISION, SHOW FILE, SEND /NOBACKUP, SET SEND
   BACKUP, and DIRECTORY /[NO]BACKUP.

4.6. Starting the Remote Kermit Server Automatically

   As noted on pages 275-276 ofUsing C-Kermit 2nd edition, you can
   have Kermit send "kermit receive" commands automatically when it is in
   local mode and you give a SEND or similar command, to start the remote
   Kermit receiver in case it is not already started. The "kermit receive"
   commands are specified by:

  SET PROTOCOL KERMIT binary-receive-command text-receive-command

   As of version 7.0, a Kermit protocol option has been added to send a
   string to the host in advance of any Kermit packets when you give a
   GET-class or REMOTE command. This will switch the remote C-Kermit into
   the appropriate mode or, if the remote system is at a system command
   (shell) prompt, execute the string on the remote system. The new syntax
   of the SET PROTOCOL KERMIT command is:

  SET PROTOCOL KERMIT [ s1 [ s2 [ s3 ] ] ]

   where:

       Default         Meaning
  s1  {kermit -ir}     Remote "kermit receive in binary mode" command.
  s2  {kermit -r}      Remote "kermit receive in text mode" command.
  s3  {kermit -x}      Remote "start kermit server" command.

   NOTE: If the remote Kermit is 6.0, the following are recommended for
   fast startup and high-performance file transfer (see Appendix I in
  Using C-Kermit, second Edition, for command-line options):

  s1   kermit -YQir   (Kermit receive binary, skip init file, fast.)
  s2   kermit -YQTr   (Kermit receive text, skip init file, fast.)
  s3   kermit -YQx    (Kermit server, skip init file, fast.)

   If the remote is C-Kermit 7.0 or later, change the -x option (enter
   server mode) to -O (uppercase letter O), which means "enter server mode
   for One transaction only); this way, it is not stuck in server after
   the transfer. Also note that the Q is redundant in version 7.0, since
   fast Kermit protocol settings are now the default.

   Note that in case the C-Kermit executable is called "wermit" or
   "ckermit" you can change "kermit" in the strings above to "wermit" or
   "ckermit" and C-Kermit 7.0 or later will recognize these as synonyms
   for "kermit", in case it is at its command prompt when one of these
   strings is sent to it.

4.7. File-Transfer Command Switches

   Over the years, various new methods of transferring a file have
   accumulated, until we had, in addition to the SEND command, also MOVE
   (send and then delete), MAIL (send as email), REMOTE PRINT (send to be
   printed), CSEND (send the output of a command), PSEND (send a part of a
   file), BSEND (send in binary mode), RESEND (resume an interrupted
   SEND), etc etc. Similarly: GET, REGET, CGET, RETRIEVE, and so on.

   Not only is it confusing to have different names for these commands,
   many of which are not real words, but this also does not allow all
   combinations, like "send a file as mail, then delete it".

   In C-Kermit 7.0, the SEND, GET, and RECEIVE commands were restructured
   to accept modifier switches (switches are explained inSection
   1.5).

4.7.1. SEND Command Switches

   Without switches, the SEND command still works exactly as before:

  send oofa.txt      ; send a single file
  send oofa.*        ; send multiple files
  send oofa.txt x.x  ; send oofa.txt as x.x (tell receiver its name is x.x)
  send               ; send from SEND-LIST

   But now the following modifier switches may be included between "send"
   and the filename. Zero, one, two, or more switches may be included in
   any combination that makes sense. Switch names (such as /BINARY) can be
   abbreviated, just like any other keywords. Most of these switches work
   only when using Kermit protocol (/TEXT and /BINARY are the exceptions).

   /AFTER:date-time
          Specifies that only those files modified (or, in VMS, created)
          after the given date-time (see Section 1.6) are to be sent.
          Examples:

  send /text /after:{2-Feb-1997 10:28:30} *.txt
  send /text /after:\fdate(oofa.txt) *.txt

          Synonym: /SINCE.

   /ARRAY:arrayname
          Specifies that instead of sending a file, C-Kermit is to send
          the contents of the given array. Since an array does not have a
          filename, you should include an /AS-NAME switch to specify the
          name under which the array is to be sent (if you do not, the
          name "_array_x_" is used, where 'x' is replaced by the array
          designator). Seesection 7.10 for array-name syntax. As
          noted in that section, you can also include a range to have a
          segment of the array sent, rather than the whole thing; for
          example: "send /array:&a[100:199]". It is strongly recommended
          that you accompany the /ARRAY switch with a /TEXT or /BINARY
          switch to force the desired transfer mode, since otherwise the
          various automatic mechanisms might switch to binary mode when
          you really wanted text, or vice versa. In text mode a line
          terminator is added to the end of each array element, but not in
          binary mode. For details and examples see Section 7.10.11.

   /AS-NAME:text
          Specifies "text" as the name to send the file under. You can
          also still specify the as-name as the second filename on the
          SEND command line. The following two commands are equivalent:

  send oofa.txt oofa.new
  send /as:oofa.new oofa.txt

   /BEFORE:date-time
          Specifies that only those files modified (or, in VMS, created)
          before the given date-time ( Section 1.6) are to be sent.

   /BINARY
          Performs this transfer in binary mode without affecting the
          global transfer mode, overriding not only the FILE TYPE and
          TRANSFER MODE settings, but also the FILE PATTERN setting, but
          for this SEND command only. In other words, SEND /BINARY means
          what it says: send the file in binary mode, regardless of any
          other settings. Example:

  set file type text      ; Set global transfer mode to text
  send /binary oofa.zip   ; Send a file in binary
  send oofa.txt           ; This one is sent in text mode

   /COMMAND
          SEND /COMMAND is equivalent to CSEND ( Section 4.2.2) -- it
          says to send the output from a command, rather than the contents
          of a file. The first "filename" on the SEND command line is
          interpreted as the name of a command; the second (if any) is the
          as-name. Examples:

  send /command {grep Sunday oofa.txt} sunday.txt
  send /as-name:sunday.txt /command {grep Sunday oofa.txt}
  send /bin /command {tar cf - . | gzip -c} {!gunzip -c | tar xf -}

   /DELETE
          Deletes the file (or each file in the group) after it has been
          sent successfully (but does not delete it if it was not sent
          successfully). SEND /DELETE is equivalent to MOVE. Has no effect
          when used with /COMMAND. Example:

  send /delete *.log

   /DOTFILES
          (UNIX and OS-9 only) Normally files whose names begin with "."
          are skipped when matching wildcards that do not also begin with
          ".". Include /DOTFILES to force these files to be included too.

   /RECURSIVE
          Descend the through the directory tree when locating files to
          send. Automatically sets /PATHNAMES:RELATIVE. Explained in
         Section 4.11 .

   /EXCEPT:pattern
          See Section 1.5.4.

   /NOBACKUP
          This means to skip backup files when sending, even if they match
          the SEND file specification. This is equivalent to using SEND
          /EXCEPT and including *.~[0-9]*~ in the exception list (or *.~*~
          if Kermit was built without pattern-matching support; see
         Section 4.9.1). Including this switch is equivalent to
          giving SET SEND BACKUP OFF ( Section 4.0.6) prior to SEND,
          except its effect is local to the SEND command with which it was
          given.

   /NODOTFILES
          The opposite of /DOTFILES (q.v.)

   /FILENAMES:{CONVERTED,LITERAL}
          Use this switch to override the current global SET FILE NAMES
          setting for this transfer only.

   /FILTER:command
          This specifies a filter to pass the file through before sending
          it. See thesection on file-transfer pipes and filters. The
          /FILTER switch applies only to the file-transfer command it is
          given with; it does not affect the global SEND FILTER setting,
          if any.

   /IMAGE
          VMS: Sends in image mode. Non-VMS: same as /BINARY.

   /LABELED
          VMS and OS/2 only: Sends in labeled mode.

   /LARGER-THAN:number
          Specifies that only those files that are longer than the given
          number of bytes are to be sent.

   /LISTFILE:filename
          Specifies that the files to be sent are listed in a file with
          the given filename. The file contains one filename per line.
          These filenames are not checked in any way; each filename is
          taken and does not use or depend on any Kermit-specific syntax.
          In particular, backslashes are not treated specially, leading
          and trailing spaces are not stripped, etc. However, if a
          filename contains wildcards, they are expanded. Example: If a
          file named files.txt contains the following lines:

  blah.txt
  oofa*
  x.x

          (but without leading or trailing spaces), then the C-Kermit
          command "send /listfile:files.txt" will send the files blah.txt,
          x.x, and all files whose names start with "oofa", assuming the
          files exist and are readable. The /LISTFILE switch, can, of
          course, be used with other switches when it makes sense, for
          example, /EXCEPT, /BINARY, /AFTER, /SMALLER, /MOVE-TO, /DELETE,
          /AS-NAME with a template, etc.

   /MAIL:address
          Sends the file as e-mail to the given address or addresses.
          "send /mail:address filename" is equivalent to "mail filename
          address". You can include multiple addresses separated by
          commas. Examples:

  send /mail:kermit-support@columbia.edu packet.log
  send /mail:cmg,fdc,jrd oofa.txt

          As with any switch argument, if the address or address list
          contains any spaces, you must enclose it in braces. The format
          of the addresses must agree with that understood by the
          mail-sending program on the receiver's computer.

   /MOVE-TO:directory-name
          Specifies that after each (or the only) source file is sent
          successfully, and ONLY if it is sent successfully, it should be
          moved to the named directory. If the directory name contains
          spaces, enclose it in braces. If the directory does not exist,
          it is created if possible; if it can't be created, the command
          fails and an error message is printed. Example:

  send /text /move-to:/users/olga/backup/ *.txt

   /NOT-AFTER:date-time
          Specifies that only those files modified at or before the given
          date and time are to be sent.

   /NOT-BEFORE:date-time
          Specifies that only those files modified at or after the given
          date and time are to be sent.

   /PATHNAMES:{OFF,ABSOLUTE,RELATIVE}
          Use this switch to override the current global SET SEND
          PATHNAMES setting for this transfer only. /PATHNAMES:ABSOLUTE or
          RELATIVE also sets /FILENAMES:LITERAL (also for this transfer
          only) since pathnames are not sent otherwise.

   /RENAME-TO:text
          Specifies that after the (or each) source file is sent
          successfully, and ONLY if it is sent successfully, it should be
          renamed to the name given. If the name contains spaces, enclose
          it in braces. If a file group is being sent, then the "text"
          must contain a variable reference such as \v(filename) (see
         Section 4.1). Example:

  send /rename-to:ok_\v(filename) *.*

          This sends each file in the current directory and if it was sent
          successfully, changes its name to begin with "ok_".

   /SMALLER-THAN:number
          Specifies that only those files that are smaller than the given
          number of bytes are to be sent.

   /SUBJECT:text
          Subject for email. Actually, this is just a synonym for
          /AS-NAME. If the text includes spaces, you must enclose it in
          braces. If you don't specify a subject (or as-name), the name of
          the file is used as the subject. Example:

  send /mail:kermit-support@columbia.edu /subj:{As requested} packet.log

   /PRINT:options
          Sends the file to be printed, optionally specifying options for
          the printer. Equivalent to REMOTE PRINT filename options.
          Examples:

  send /print oofa.txt              ; No options.
  send /print:/copies=3 oofa.txt    ; "/copies=3" is a VMS PRINT switch.
  send /print:-#3 oofa.txt          ; "-#3" is a UNIX lpr switch.

   /PROTOCOL:name
          Uses the given protocol to send the file (Kermit, Zmodem, etc)
          for this transfer without changing global protocol. Only
          available in Kermit 95, UNIX, and OS-9. Example:

  set protocol kermit               ; Set global protocol
  send /proto:zmodem /bin oofa.zip  ; Send just this file with Zmodem
  send oofa.txt                     ; This file is sent with Kermit

   /QUIET
          When sending in local mode, this suppresses the file-transfer
          display.

   /RECOVER
          Used to recover from a previously interrupted transfer; SEND
          /RECOVER is equivalent to RESEND. Recovery only works in binary
          mode; SEND /RECOVER and RESEND include an implied /BINARY
          switch. Even then, recovery will successful only if (a) the
          original (interrupted) transfer was also in binary mode, or (b)
          if it was in text mode, the two Kermit programs run on platforms
          where text-mode transfers are not length-changing.

   /STARTING:number
          Starts sending the file from the given byte position. SEND
          /STARTING:n filename is equivalent to PSEND filename n.

   /TEXT
          Performs this transfer in text mode without affecting the global
          transfer mode, overriding not only the FILE TYPE and TRANSFER
          MODE settings, but also the FILE PATTERN setting, for this SEND
          command only. In other words, SEND /TEXT really send the file in
          text mode, regardless of any other settings or negotiations.

   About mail... Refer to Section 4.7.1. The same rules apply as for
   file transfer. If you are mailing multiple files, you can't use an
   as-name (in this case, a subject) unless it contains replacement
   variables like \v(filenum). For example, if you:

  send /mail:somebody@xyz.com *.txt

   Then each file will arrive as a separate email message with its name as
   the subject. But if you:

  send /mail:somebody@xyz.com /subject:{Here is a file} *.txt

   Then each file message will have the same subject, which is probably
   not what you want. You can get around this with constructions like:

  send /mail:somebody@xyz.com /subject:{Here is \v(filename)} *.txt

   which embed the filename in the subject.

   The MOVE, CSEND, MAIL, and RESEND commands now also accept the same
   switches. And the switches are also operative when sending from a
   SEND-LIST (seeUsing C-Kermit, 2nd Ed, pp.191-192), so, for
   example, it is now possible to SEND /PRINT or SEND /MAIL from a
   SEND-LIST.

   The MSEND and MMOVE commands also take switches, but not all of them.
   With these commands, which take an arbitrary list of filespecs, you can
   use /BINARY, /DELETE, /MAIL, /PRINT, /PROTOCOL, /QUIET, /RECOVER, and
   /TEXT (and /IMAGE or /LABELED, depending on the platform). MMOVE is
   equivalent to MSEND /DELETE. (If you want to send a group of files, but
   in mixed transfer modes with per-file as-names, use ADD SEND-LIST and
   then SEND.)

   The MSEND/MMOVE switches come before the filenames, and apply to all of
   them:

  msend /print /text *.log oofa.txt /etc/motd

   If you type any of these commands (SEND, CSEND, MSEND, etc) followed by
   a question mark (?), you will see a list of the switches you can use.
   If you want to see a list of filenames, you'll need to type something
   like "send ./?" (UNIX, OS/2, Windows, etc), or "send []?" (VMS), etc.
   Of course, you can also type pieces of a filename (anything that does
   not start with "/") and then "?" to get a list of filenames that start
   that way; e.g. "send x.?" still works as before.

   In UNIX, where "/" is also the directory separator, there is usually no
   ambiguity between a fully-specified pathname and a switch, except when
   a file in the root directory has the same name as a switch (as noted in
  Section 1.5):

  send /etc/motd                        ; Works as expected
  send /command                         ; ???

   The second example interprets "/command" as a switch, not a filename.
   To send a file actually called "command" in the root directory, use:

  send {/command}

   or other system-dependent forms such as //command, /./command,
   c:/command, etc, or cd to / and then "send command".

4.7.2. GET Command Switches

   Without switches, the GET command still works about the same as before:

  get oofa.txt                          ; GET a single file
  get oofa.*                            ; GET multiple files

   However, the mechanism for including an "as-name" has changed.
   Previously, in order to include an as-name, you were required to use
   the "multiline" form of GET:

  get
  remote-filespec
  local-name

   This was because the remote filespec might contain spaces, and so there
   would be no good way of telling where it ended and where the local name
   began, e.g:

  get profile exec a foo

   But now since we can use {braces} for grouping, we don't need the
   multiline GET form any more, and in fact, support for it has been
   removed. If you give a GET command by itself on a line, it fails and an
   error message is printed. The new form is:

   GET [ switches... ] remote-name [ local-name ]
          Ask the server to send the file whose name is remote-name. If
          the optional local-name is given, store it locally under this
          name. If the remote-name or local-name contains spaces, they
          must be enclosed in braces:

  get {profile exec a} foo
  get oofa.txt {~/My Files/Oofa text}

   If you want to give a list of remote file specifications, use the MGET
   command:

   MGET [ switches... ] remote-name [ remote-name [ remote-name ... ] ]
          Ask the server to send the files whose names are given.

   Now you can also include modifier switches between GET or MGET and the
   remote-name; most of the same switches as SEND:

   /AS-NAME:text
          Specifies "text" as the name to store the incoming file under.
          (This switch is not available for MGET.) You can also still
          specify the as-name as the second filename on the GET command
          line. The following two commands are equivalent:

  get oofa.txt oofa.new
  get /as:oofa.new oofa.txt

   /BINARY
          Tells the server to send the given file(s) in binary mode
          without affecting the global transfer mode. Example:

  set file type text      ; Set global transfer mode to text
  get /binary oofa.zip    ; get a file in binary mode
  get oofa.txt            ; This one is transferred in text mode

          Or, perhaps more to the point:

  get /binary foo.txt     ; where "*.txt" is a text-pattern

          This has the expected effect only if the server is C-Kermit 7.0
          or later or K95 1.1.19 or later.

   /COMMAND
          GET /COMMAND is equivalent to CGET ( Section 4.2.2) -- it
          says to receive the file into the standard input of a command,
          rather than saving it on disk. The /AS-NAME or the second
          "filename" on the GET command line is interpreted as the name of
          a command. Examples:

  get /command sunday.txt {grep Sunday oofa.txt}
  get /command /as-name:{grep Sunday oofa.txt} sunday.txt
  get /bin /command {!gunzip -c | tar xf -} {tar cf - . | gzip -c}

   /DELETE
          Asks the Kermit server to delete the file (or each file in the
          group) after it has been transferred successfully (but not to
          delete it if it was not sent successfully). GET /DELETE is
          equivalent to RETRIEVE. Example:

  get /delete *.log

   /EXCEPT:pattern
          Specifies that any files whose names match the pattern, which
          can be a regular filename, or may contain "*" and/or "?"
          metacharacters, are to be refused upon arrival. To specify
          multiple patterns (up to 8), use outer braces around the group,
          and inner braces around each pattern:

  /EXCEPT:{{pattern1}{pattern2}...}

          See the description of SEND /EXCEPT in Section 4.7.1 for
          examples, etc. Refusal is accomplished using the Attribute
          Rejection mechanism (reason "name"), which works only when
          Attribute packets have been successfully negotiated.

   /FILENAMES:{CONVERTED,LITERAL}
          Use this switch to override the current global SET FILE NAMES
          setting for this transfer only.

   /FILTER:command
          This specifies a filter to pass the incoming file through before
          writing to disk. See thesection on file-transfer pipes and
          filters. The /FILTER switch applies only to the file-transfer
          command it is given with; it does not affect the global RECEIVE
          FILTER setting, if any.

   /IMAGE
          VMS: Transfer in image mode. Non-VMS: same as /BINARY.

   /LABELED
          VMS and OS/2 only: Specifies labeled transfer mode.

   /MOVE-TO:directory
          This tells C-Kermit to move each file that is successfully
          received to the given directory. Files that are not successfully
          received are not moved. By default, files are not moved.

   /PATHNAMES:{OFF,ABSOLUTE,RELATIVE,AUTO}
          Use this switch to override the current global SET RECEIVE
          PATHNAMES setting for this transfer only. /PATHNAMES:ABSOLUTE or
          RELATIVE also sets /FILENAMES:LITERAL (also for this transfer
          only) since incoming pathnames would not be treated as pathnames
          otherwise. See Section 4.10.

   /QUIET
          When sending in local mode, this suppresses the file-transfer
          display.

   /RECOVER
          Used to recover from a previously interrupted transfer; GET
          /RECOVER is equivalent to REGET. Recovery only works in binary
          mode; SEND /RECOVER and RESEND include an implied /BINARY
          switch. Even then, recovery will successful only if (a) the
          original (interrupted) transfer was also in binary mode, or (b)
          if it was in text mode, the two Kermit programs run on platforms
          where text-mode transfers are not length-changing.

   /RECURSIVE
          Tells the server that the GET file specification applies
          recursively. This switch also automatically sets
          /PATHNAMES:RELATIVE in both the server AND the client. When used
          in conjunction with /DELETE, this "moves" a directory tree from
          the server's computer to the client's computer (except that only
          regular files are deleted from the server's computer, not
          directories; thus the original directories will be left, but
          will contain no files). Note that all servers that support
          /RECURSIVE do not necessarily do so in combination with other
          switches, such as /RECOVER. (Servers that do include C-Kermit
          7.0 and later, K95 1.1.19 and later.)

   /RENAME-TO:string
          This tells C-Kermit to rename each file that is successfully
          received to the given string. Files that are not successfully
          received are not renamed. By default, files are not renamed. The
          string can be a literal string, which is appropriate when only
          one file is being received, or it can contain one or more
          variables that are to be evaluated at the time each file is
          received, such as \v(filename), \v(filenumber), \v(ntime),
          \v(pid), \v(user), etc. WARNING: if you give a literal string
          and more than one file arrives, each incoming file will be given
          the same name (but SET FILE COLLISION BACKUP or RENAME can be
          used to keep the incoming files from overwriting each other).

   /TEXT
          Tells the server to perform this transfer in text mode without
          affecting its global transfer mode. See /BINARY for additional
          info.

   The /MAIL and /PRINT options are not available (as they are for SEND),
   but you can use /COMMAND to achieve the same effect, as in these UNIX
   examples:

  get /command oofa.txt {mail kermit@columbia.edu}
  get /command oofa.txt lpr

   In OS/2 or Windows, you can GET and print like this:

  get oofa.txt prn

   The CGET, REGET, RETRIEVE commands also accept the same switches as
   GET. CGET automatically sets /COMMAND; REGET automatically sets
   /RECOVER and /BINARY, and RETRIEVE automatically sets /DELETE.

4.7.3. RECEIVE Command Switches

   Without switches, the RECEIVE command still works as before:

  receive            ; Receives files under their own names
  receive /tmp       ; Ditto, but into the /tmp directory
  r                  ; Same as "receive"
  receive foo.txt    ; Receives a file and renames to foo.txt

   Now you can also include modifier switches may be included between
   "receive" and the as-name; most of the same switches as GET:

   /AS-NAME:text
          Specifies "text" as the name to store the incoming file under.
          You can also still specify the as-name as a filename on the
          command line. The following two commands are equivalent:

  r oofa.new
  r /as:oofa.new

   /BINARY
          Performs this transfer in binary mode without affecting the
          global transfer mode. NOTE: This does not override the incoming
          filetype (as it does with GET), so this switch is useful only if
          ATTRIBUTE TYPE is OFF, or if the other Kermit does not send a
          TYPE (text or binary) attribute. In any case, it has no affect
          whatsoever on the file sender.

   /COMMAND
          RECEIVE /COMMAND is equivalent to CRECEIVE ( Section 4.2.2)
          -- it says to receive the file into the standard input of a
          command, rather than saving it on disk. The /AS-NAME or the
          "filename" on the RECEIVE command line is interpreted as the
          name of a command.

  r /command {grep Sunday oofa.txt}
  r /command /as-name:{grep Sunday oofa.txt}
  r /bin /command {tar cf - . | gzip -c}

   /EXCEPT:pattern
          Specifies that any files whose names match the pattern, which
          can be a regular filename, or may contain "*" and/or "?"
          metacharacters, are to be refused upon arrival. To specify
          multiple patterns (up to 8), use outer braces around the group,
          and inner braces around each pattern:

  /EXCEPT:{{pattern1}{pattern2}...}

          See the description of SEND /EXCEPT in Section 4.7.1 for
          examples, etc. Refusal is accomplished using the Attribute
          Rejection mechanism (reason "name"), which works only when
          Attribute packets have been successfully negotiated.

   /FILENAMES:{CONVERTED,LITERAL}
          Use this switch to override the current global SET FILE NAMES
          setting for this transfer only.

   /FILTER:command
          This specifies a filter to pass the incoming file through before
          writing to disk. See thesection on file-transfer pipes and
          filters. The /FILTER switch applies only to the file-transfer
          command it is given with; it does not affect the global RECEIVE
          FILTER setting, if any.

   /IMAGE
          VMS: Transfer in image mode. Non-VMS: same as /BINARY. See
          comments under RECEIVE /BINARY.

   /LABELED
          VMS and OS/2 only: Specifies labeled transfer mode. See comments
          under RECEIVE /BINARY.

   /MOVE-TO:directory
          This tells C-Kermit to move each file that is successfully
          received to the given directory. Files that are not successfully
          received are not moved. By default, files are not moved.

   /PATHNAMES:{ABSOLUTE,RELATIVE,OFF,AUTO}
          Use this switch to override the current global SET RECEIVE
          PATHNAMES setting for this transfer only. See Section 4.10.

   /RECURSIVE
          When used with the RECEIVE command, /RECURSIVE is simply a
          synonym for /PATHNAMES:RELATIVE.

   /RENAME-TO:string
          This tells C-Kermit to rename each file that is successfully
          received to the given string. Files that are not successfully
          received are not renamed. By default, files are not renamed. The
          string can be a literal string, which is appropriate when only
          one file is being received, or it can contain one or more
          variables that are to be evaluated at the time each file is
          received, such as \v(filename), \v(filenumber), \v(ntime),
          \v(pid), \v(user), etc. WARNING: if you give a literal string
          and more than one file arrives, each incoming file will be given
          the same name (but SET FILE COLLISION BACKUP or RENAME can be
          used to keep the incoming files from overwriting each other).

   /QUIET
          When receiving in local mode, this suppresses the file-transfer
          display.

   /TEXT
          Receives in text mode without affecting the global transfer
          mode. See comments under RECEIVE /BINARY.

   The /MAIL and /PRINT options are not available, but you can use
   /COMMAND to achieve the same effect, as in these UNIX examples:

  r /command {mail kermit@columbia.edu}
  r /command lpr

   In OS/2 or Windows, you can RECEIVE and print like this:

  receive prn

   The CRECEIVE command now also accepts the same switches.

4.8. Minor Kermit Protocol Improvements

4.8.1. Multiple Attribute Packets

   C-Kermit 7.0 now sends more than one Attribute packet if a file's
   attributes do not fit into a single packet of the negotiated length. If
   a particular attribute (such as file creation date-time) does not fit
   within the negotiated length (which will only happen when the
   negotiated length is around 20 or less), that attribute is not sent at
   all.

4.8.2. Very Short Packets

   There are certain situations where extremely short packets must be
   used; 20 or 30 bytes at most. This can happen when one or more devices
   along the communication path have very small buffers and lack an
   effective means of flow control. Examples are sometimes cited involving
   radio modems.

   When the maximum packet length is shorter than certain packets that
   would be sent, those packets are either truncated or else broken up
   into multiple packets. Specifically:

    1. Parameter negotiation packets (I, S, and their ACKs) are truncated
       to the negotiated length. Any parameters that do not fit are reset
       to their default values. There is no provision in the Kermit
       protocol for fragmentation and reassembly of parameter strings.
    2. File header packets (containing the filename) are simply truncated.
       There is no provision in the Kermit protocol for fragmentation and
       reassembly of filenames.
    3. Attribute packets are fragmented and reassembled as described in
       4.8.1 without loss of data, except in case a field will not fit at
       all in the negotiated length (the longest attribute is usually the
       date and time of file creation/modification) because of the rule
       that attributes may not be broken across packets.
    4. Data packets and other packets are unaffected -- they can be as
       short as they need to be, within reason.

4.9. Wildcard / File Group Expansion

   "Wildcard" refers to the notation used in filenames to specify a group
   of files by pattern matching.

4.9.1. In UNIX C-Kermit

   Prior to C-Kermit 7.0, C-Kermit was capable of expanding wildcard
   strings containing only the "metacharacters" '*' and '?':

   *
          Matches any sequence of zero or more characters. For example:
          "ck*.c" matches all files whose names start with "ck" and end
          with ".c", including "ck.c".

   ?
          Matches any single character. For example, "ck?.c" matches all
          files whose names are exactly 5 characters long and start with
          "ck" and end with ".c". When typing commands at the prompt, you
          must precede any question mark to be used for matching by a
          backslash (\) to override the normal function of question mark,
          which is providing menus and file lists.

   C-Kermit 7.0 adds the additional features that users of ksh, csh, and
   bash are accustomed to:

   [abc]
          Square brackets enclosing a list of characters matches any
          single character in the list. Example: ckuusr.[ch] matches
          ckuusr.c and ckuusr.h.

   [a-z]
          Square brackets enclosing a range of characters; the hyphen
          separates the low and high elements of the range. For example,
          [a-z] matches any character from a to z.

   [acdm-z]
          Lists and ranges may be combined. This example matches a, c, d,
          or m through z.

   {string1,string2,...}
          Braces enclose a list of strings to be matched. For example:
          ck{ufio,vcon,cmai}.c matches ckufio.c, ckvcon.c, or ckcmai.c.
          The strings may themselves contain metacharacters, bracket
          lists, or indeed, other lists of strings, but (when matching
          filenames) they may not contain directory separators.

          Thus, the metacharacters in filenames (and in any other field
          that can be a pattern, such as the IF MATCH pattern, SEND or GET
          exception lists, etc) are:

 * ? [ {

          And within braces only, comma (,) is a metacharacter.

   To include a metacharacter in a pattern literally, precede it with a
   backslash '\' (or two if you are passing the pattern to a macro).
   Examples:

  send a*b      ; Send all files whose names start with 'a' and end with 'b'.
  send a?b      ; Ditto, but the name must be exactly three characters long.
  send a[a-z]b  ; Ditto, but the second character must be a lowercase letter.
  send a[x\-z]b ; Ditto, except the second character must be 'x', '-', or 'y'.
  send a[ghi]b  ; Ditto, except the second character must be 'g', 'h', or 'i'.
  send a[?*]b   ; Ditto, except the second character must be '?' or '*'.
  send a[\?\*]b ; Same as previous.
  send *?[a-z]* ; All files with names containing at least one character
                ; that is followed by a lowercase letter.

   Or, more practically:

  send ck[cuw]*.[cwh]  ; Send the UNIX C-Kermit source files.

   To refer to the C-Kermit sources files and makefile all in one
   filespec:

  {{makefile,ck[cuw]*.[cwh]}}

   (NOTE: if the entire pattern is a {stringlist}, you must enclose it it
   TWO pairs of braces, since the SEND command strips the outer brace
   pair, because of the "enclose in braces if the filename contains
   spaces" rule).

   If the makefile is called ckuker.mak:

  ck[cuw]*.{[cwh],mak}

   (NOTE: double braces are not needed here since the pattern does not
   both begin and end with a brace.)

   To add in all the C-Kermit text files:

  ck[cuw]*.{[cwh],mak,txt}

   All of these features can be used anywhere you would type a filename
   that is allowed to contain wildcards.

   When you are typing at the command prompt, an extra level of quoting is
   required for the '?' character to defeat its regular function of
   producing a list of files that match what you have typed so far, for
   example:

  send ck[cu]?

   lists all the files whose names start with ckc and cku. If you quote
   the question mark, it is used as a pattern-matching character, for
   example:

  send ck\?[ft]io.c

   sends all the file and communications i/o modules for all the
   platforms: ckufio.c, ckutio.c, ckvfio.c, ckvtio.c, etc.

   If, however, a filename actually contains a question mark and you need
   to refer to it on the command line, you must use three (3) backslashes.
   For example, if the file is actually called ck?fio.c, you would use:

  send ck\\\?fio.c

   Further notes on quoting:

     * A single backslash is sufficient for quoting a special character at
       the command prompt or in a command file. However, when passing
       patterns to macros you'll need double backslashes, and when
       referring to these patterns within the macro, you'll need to use
       \fcontents(\%1) (see Section 1.11.5). You should enclose macro
       argument references in braces in case grouped arguments were
       passed. Example:
 define ismatch {
     if match {\fcont(\%1)} {\fcont(\%2)} {
         end 0 MATCH
     } else {
         end 1 NO MATCH
     }
 }
 ismatch ab*yz a*\\**z           ; Backslash must be doubled
 ismatch {abc def xyz} *b*e*y*   ; Braces must be used for grouping

     * Watch out for possible conflicts between {} in filename patterns
       and {} used for grouping multiple words into a single field, when
       the pattern has outer braces. For example, in:
     if match {abc xyz} {a* *z} echo THEY MATCH

       braces must be used to group "abc xyz" into a single string. Kermit
       strips off the braces before comparing the string with the pattern.
       Therefore:
 if match makefile {makefile,Makefile} echo THEY MATCH

       does not work, but:
 if match makefile {{makefile,Makefile}} echo THEY MATCH

       does.
     * If you use a pattern that has outer braces, like {*.txt,*.doc}, in
       a field that accepts a pattern list (like SEND /EXCEPT:xxx), you'll
       need to add two extra sets of outer braces:
 send /except:{{{*.txt,*.doc}}} *.*

   C-Kermit's new pattern matching capabilities are also used when
   C-Kermit is in server mode, so now you can send requests such as:

  get ck[cuw]*.[cwh]

   to a C-Kermit server without having to tell it to SET WILD SHELL first.
   Previously this would have required:

  mget ckc*.c ckc*.w ckc*.h cku*.c cku*.w cku*.h ckw*.c ckw*.w ckw*.h

   The new pattern matching features make SET WILD SHELL redundant, and
   barring any objections, it will eventually be phased out. (One possible
   reason for retaining it would be as an escape mechanism when Kermit
   does not understand the underlying file system.)

   By the way, patterns such as these are sometimes referred to as
   "regular expressions", but they are not quite the same. In a true
   regular expression (for example), "*" means "zero or more repetitions
   of the previous item", so (for example), "([0-9]*)" would match zero or
   more digits in parentheses. In Kermit (and in most shells), this
   matches one digit followed by zero or more characters, within
   parentheses. Here are some hints:

     * Although you can't match any sequence of digits (or letters, etc),
       you can match (say) 1, 2, or 3 of them in row. For example, the
       following pattern matches Kermit backup files (with backup numbers
       from 1 to 999):
 *.~{[1-9],[1-9][0-9],[1-9][0-9][0-9]}~

     * There is presently no NOT operator, so no way to match any
       character or string EXCEPT the one(s) shown.

   In other wildcarding news...

     * You may now "send xxx" where "xxx" is a directory name, and this
       will send all the files from the directory xxx, as if you had typed
       "send xxx/*". You can also use the special shorthand "send ." to
       send all the files from the current directory.
     * To easily skip over backup files (the ones whose names end like
       .~22~) when sending, you can use SEND /NOBACKUP (seeSection
       4.0.6 for details).
     * When choosing Kermit to expand wildcards, rather than the shell,
       you can choose whether "dot files" -- files whose names begin with
       ".", which are normally "invisible" -- should be matched:
 SET WILD KERMIT /NO-MATCH-DOT-FILES (this is the default)
 SET WILD KERMIT /MATCH-DOT-FILES    (this allows matching of "." files)

       or include the /DOTFILES or /NODOTFILES switch on the command you
       are using, such as SEND or DIRECTORY.
     * Commands such as DIRECTORY and SEND allow recursive directory
       traversal. There are also new functions for this to use in scripts.
       See Section 4.11 for details.

   When building file lists in UNIX, C-Kermit follows symbolic links.
   Because of this, you might encounter any or all of the following
   phenomena:

     * Multiple copies of the same file; e.g. one from its real directory
       and others from links to its real directory, if both the real
       directory and the links to it are in the wildcard expansion list.
     * A command might unexpectedly "hang" for a long time because an NFS
       link might not be responding, or the directory you are looking at
       contains a link to a huge directory tree (example: "directory
       /recursive /etc" when /etc/spool is a symlink to /var/spool, which
       is a large organization's incoming email directory, containing tens
       of thousands of subdirectories).

   The size of the file list that Kermit can build is limited in most
   C-Kermit implementations. The limit, if any, depends on the
   implementation. Use the SHOW FEATURES command and look in the
   alphabetized options list for MAXWLD to see the value.

4.9.2. In Kermit 95

   Kermit 95 1.1.19 and later uses the same pattern matching syntax as in
   UNIX, but (as always) you will encounter numerous difficulties if you
   use backslash (\) as the directory separator. In any command where K95
   parses filenames itself (that is, practically any file-oriented command
   except RUN), you can use forward slash (/) as the directory separator
   to avoid all the nasty conflicts.

4.9.3. In VMS, AOS/VS, OS-9, VOS, etc.

   Platforms other than UNIX, Windows 95/98/NT, and OS/2 have their own
   filename matching capabilities that are, in general, different from
   Kermit's built-in ones and in any case might conflict with them. For
   example, [] encloses directory names in VMS.

   Nevertheless you can still use all the pattern-matching capabilities
   described in Section 4.9.1 by loading a file list into an array
   (e.g. with \ffiles(*,&a), see Section 4.11.3) and then using IF
   MATCH on the members.

4.10. Additional Pathname Controls

   In version 6.0 and earlier, C-Kermit's SET { SEND, RECEIVE } PATHNAMES
   command had only ON and OFF as options. In version 7.0, there are more
   choices:

   SET SEND PATHNAMES OFF
          When sending a file, strip all disk/directory information from
          the name. Example: "send /usr/olga/letters/oofa.txt" sends the
          file as "oofa.txt". This applies to actual filenames, not to any
          as-name you might specify.

   SET SEND PATHNAMES RELATIVE
          When sending a file, leave the pathname on as given. For
          example, if your current directory is /usr/olga, "send
          letters/oofa.txt" sends the file as "letters/oofa.txt", not
          "/usr/olga/letters/oofa.txt" or "letters.txt".

   SET SEND PATHNAMES ABSOLUTE
          When sending a file, convert its name to the full, absolute
          local pathname. For example, if your current directory is
          /usr/olga, "send letters/oofa.txt" sends the file as
          "/usr/olga/letters/oofa.txt". NOTE: Even with this setting,
          device and/or node names are not included. For example, in VMS,
          any node or device name is stripped; in Windows or OS/2, any
          disk letter is stripped.

   SET RECEIVE PATHNAMES OFF
          When receiving a file, strip all disk/directory information from
          the name before attempting to store it. This applies to incoming
          filename, not to any as-name you might specify. Example: If a
          file arrives under the name "/usr/olga/letters/oofa.txt" it is
          stored simply as "oofa.txt" in your download directory or, if no
          download directory has been specified, in your current
          directory.

   SET RECEIVE PATHNAMES RELATIVE
          When receiving a file, leave the pathname on as it appears in
          the incoming name, but if the incoming name appears to be
          absolute, make it relative to your current or download
          directory. Examples:

          + "oofa.txt" is stored as "oofa.txt".
          + "letters/oofa.txt" is stored as "letters/oofa.txt"; the
            "letters" subdirectory is created if it does not already
            exist.
          + "/usr/olga/letters/oofa.txt" is stored as
            "usr/olga/letters/oofa.txt" in your current or download
            directory, and the "usr", "usr/olga", etc, directories are
            created if they do not exist.

   SET RECEIVE PATHNAMES ABSOLUTE
          The incoming filename is used as given. Thus it cannot be stored
          unless the given path (if any) already exists or can be created.
          In this case, node, device, or disk designations are NOT
          stripped, since they most likely were given explicitly by the
          user as an as-name, meant to be used as given.

   SET RECEIVE PATHNAMES AUTO
          This is the default, and means RELATIVE if the sender tells me
          it is a recursive transfer, OFF otherwise.

   Set FILE NAMES CONVERTED now also affects pathnames too. When PATHNAMES
   are RELATIVE or ABSOLUTE and FILE NAMES are CONVERTED, the file sender
   converts its native directory-name format to UNIX format, and the file
   receiver converts from UNIX format to its native one; thus UNIX format
   is the common intermediate representation for directory hierarchies, as
   it is in the ZIP/UNZIP programs (which is why ZIP archives are
   transportable among, UNIX, DOS, and VMS).

   Here's an example in which a file is sent from Windows to UNIX with
   relative pathnames and FILE NAMES CONVERTED:

  Source name                Intermediate name      Destination Name
  C:\K95\TMP\OOFA.TXT        K95/TMP/OOFA.TXT       k95/tmp/oofa.txt

   In a more complicated example, we send the same file from Windows to
   VMS:

  Source name                Intermediate name      Destination Name
  C:\K95\TMP\OOFA.TXT        K95/TMP/OOFA.TXT       [.K95.TMP]OOFA.TXT

   (Note that disk letters and device designations are always stripped
   when pathnames are relative).

   As you can imagine, as more and more directory formats are considered,
   this approach keeps matters simple: on each platform, Kermit must know
   only its own local format and the common intermediate one. In most
   cases, the receiver can detect which format is used automatically.

4.11. Recursive SEND and GET: Transferring Directory Trees

   C-Kermit 7.0 in selected versions (UNIX, VMS, VOS, AOS/VS, Windows, and
   OS/2 at this writing) now permits the SEND command to traverse
   directories "recursively" if you ask it to; that is, to send files from
   the current or specified directory and all of its subdirectories too,
   and their subdirectories, etc. (Some other commands can do this too,
   including DIRECTORY.)

   This feature is new to UNIX, Windows, VOS, and OS/2. VMS and AOS/VS
   have always included "wildcard" or "template" characters that allow
   this, and in this case, recursive directory traversal could happen
   behind Kermit's back, i.e. Kermit does not have to do it itself (in
   VMS, the notation is "[...]" or "[directory...]"; in AOS/VS is "#"). In
   C-Kermit 7.0, however, SEND /RECURSIVE is supported by C-Kermit itself
   for VMS.

4.11.1. Command-Line Options

   To descend a directory tree when sending files, use the -L command-line
   option to indicate that the send operation is to be recursive, and
   include a name or pattern to be sent. When giving a pattern, you should
   enclose it in quotes to prevent the shell from expanding it. Examples:

  $ kermit -Ls "/usr/olga/*" # send all of Olga's files in all her directories
  $ kermit -Ls foo.txt       # send all foo.txt files in this directory tree
  $ kermit -Ls "*.txt"       # send all .txt files in this directory tree
  $ kermit -Ls "letters/*"   # send all files in the letters directory tree
  $ kermit -Ls letters       # send all files in the letters directory tree
  $ kermit -Ls "*"           # send all files in this directory tree
  $ kermit -Ls .             # UNIX only: send all files in this directory tree
  $ kermit -s .              # UNIX only: a filename of . implies -L

   If you let the shell expand wildcards, Kermit only sends files whose
   names match files in the current or given directory, because the shell
   replaces an unquoted wildcard expression with the list of matching
   files -- and the shell does not build recursive lists. Note that the
   "." notation for the tree rooted at the current directory is allowed
   only in UNIX, since in Windows and OS/2, it means "*.*" (nonrecursive).

4.11.2. The SEND /RECURSIVE Command

   If you include the /RECURSIVE switch in a SEND (or MOVE, or similar)
   command, it means to descend the current or specified directory tree
   searching for files whose names match the given name or pattern. Since
   this is not terribly useful unless you also include pathnames with the
   outbound files, the /RECURSIVE switch also includes an implicit
   /PATHNAMES:RELATIVE switch (which you can undo by including an explicit
   /PATHNAMES switch after the /RECURSIVE switch).

   Examples:

   SEND /RECURSIVE *
          Sends all of the files in the current directory and all the
          files in all of its subdirectories, and all of their
          subdirectories, etc, including their relative pathnames. Empty
          directories are not sent.

   SEND /RECURSIVE /PATHNAMES:ABSOLUTE *
          Sends all of the files in the current directory and all the
          files in all of its subdirectories, and all of their
          subdirectories, etc, including their absolute pathnames.

   SEND /RECURSIVE /PATHNAMES:OFF *
          Sends all of the files in the current directory and all the
          files in all of its subdirectories, and all of their
          subdirectories, etc, without pathnames.

   SEND /RECURSIVE /usr/olga/*
          Sends all of the files in the /usr/olga directory and all the
          files in all of its subdirectories, and all of their
          subdirectories, etc.

   SEND /RECURSIVE /usr/olga (or /usr/olga/)
          Same as above. If the name is a directory name (with or without
          a trailing slash), its files are sent, and those of its
          subdirectories, and their subdirectories, etc (seeSection
          4.9).

   SEND /RECURSIVE /TEXT /usr/olga/*.txt
          As above, but only files whose names end with ".txt" are sent,
          and they are sent in text mode (as they would be by default
          anyway if SET FILE PATTERNS is ON or AUTO).

   SEND .
          UNIX only: Send all the files in the current directory.

   SEND /RECURSIVE .
          UNIX only: Sends all of the files in the current directory and
          all of its subdirectories, etc ( Section 4.9).

   The /RECURSIVE switch is different from most other switches in that its
   effect is immediate (but still local to the command in which it is
   given), because it determines how filenames are to be parsed. For
   example, "send *.txt" fails with a parse error ("No files match") if
   there are no *.txt files in the current directory, but "send /recursive
   *.txt" succeeds if there are ".txt" files anywhere in the tree rooted
   at the current directory.

   The /RECURSIVE switch also affects the file lists displayed if you type
   "?" in a filename field. "send ./?" lists the regular files in the
   current directory, but "send /recursive ./?" lists the entire directory
   tree rooted at the current directory.

4.11.3. The GET /RECURSIVE Command

   In a client/server setting, the client can also request a recursive
   transfer with:

   GET /RECURSIVE [ other switches ] remote-filespec [ local-spec ]

   In which remote file specification can be a directory name, a filename,
   a wildcard, or any combination. If the local-spec is not given (and
   PATHNAMES are RELATIVE), incoming files and directories go into the
   current local directory. If local-spec is given and is a directory, it
   becomes the root of the tree into which the incoming files and
   directories are placed. If local-spec has the syntax of a directory
   name (e.g. in UNIX it ends with /), C-Kermit creates the directory and
   then places the incoming files into it. If local-spec is a filename
   (not recommended), then all incoming files are stored with that name
   with collisions handled according to the FILE COLLISION setting.

   Again, the normal method for transferring directory trees uses relative
   pathnames, and this is the default when the sender has been given the
   /RECURSIVE switch. The action at the receiver depends on its RECEIVE
   PATHNAMES setting. The default is AUTO, meaning that if the sender
   tells it to expect a recursive transfer, then it should automatically
   switch to relative pathnames for this transfer only; otherwise it obeys
   the RECEIVE PATHNAMES setting of OFF, ABSOLUTE, or RELATIVE.

   What happens if a file arrives that has an absolute pathname, when the
   receiver has been told to use only relative pathnames? As a security
   precaution, in this case the receiver treats the name as if it was
   relative. For example, if a file arrives as:

  /usr/olga/oofa.txt

   The receiver creates a "usr" subdirectory in its current directory, and
   then an "olga" subdirectory under the "usr" subdirectory in which to
   store the incoming file.

   Suppose, however there is a sequence of directories:

  /usr/olga/a/b/c/d/

   in which "a" contains nothing but a subdirectory "b", which in turn
   contains nothing but a subdirectory "c", which in turn contains nothing
   but a subdirectory "d", which contains nothing at all. Thus there are
   no files in the "/usr/olga/a/" tree, and so it is not sent, and
   therefore it is not reproduced on the target computer.

4.11.4. New and Changed File Functions

   C-Kermit 7.0 adds the following functions:

   \ffiles(pattern[,&a])
          This function has been changed to match only regular files in
          the current or given directory, and to take an optional array
          name as a second argument (explained below).

   \fdirectories(pattern[,&a])
          Returns the number of directories that match the given pattern.
          If the pattern does not include a directory, then the search is
          performed in the current directory.

   \frfiles(pattern[,&a])
          Returns the number of files in the current or given directory
          and all of its subdirectories, and their subdirectories, etc,
          that match the given pattern. Warning -- this one can take quite
          some time if performed at the root of a large directory tree.

   \frdirectories(pattern[,&a])
          Returns the number of directories in the current or given
          directory and all of its subdirectories, and their
          subdirectories, etc, that match the given pattern.

   Each of these functions builds up a list of files to be returned by the
   \fnextfile() function, just as \ffiles() always has done. (This can
   also be done with the /ARRAY switch of the DIRECTORY command; see
 Sections 4.5.1 and7.10).

   Each of these functions can be given an array name as an optional
   second argument. If an array name is supplied, the array will contain
   the number of files as its 0th element, and the filenames in elements 1
   through last. If the array already existed, its previous contents are
   lost. For example, if the current directory contains two files,
   oofa.txt and foo.bar, then "\ffiles(*,&a)" creates an array \&a[] with
   a dimension of 2, containing the following elements:

 \&a[0] = 2
 \&a[1] = oofa.txt
 \&a[2] = foo.bar

   If no files match the specification given in the first argument, the
   array gets a dimension of 0, which is the same as undeclaring the
   array.

   Note that the order in which the array is filled (and in which
   \fnextfile() returns filenames) is indeterminate (but seeSection
   7.10.5).

   Here's an example that builds and prints a list of all the file whose
   names end in .txt in the current directory and all its descendents:

  asg \%n \frfiles(*.txt)
  declare \&a[\%n]
  for \%i 1 \%n 1 {
      asg \&a[\%i] \fnextfile()
      echo \flpad(\%i,4). "\&a[\%i]"
  }

   Alternatively, using the array method, and then printing the filenames
   in alphabetic order (see Section 7.10.3 and7.10.5):

  asg \%n \frfiles(*.txt,&a)
  sort &a
  for \%i 1 \%n 1 {
      echo \flpad(\%i,4). "\&a[\%i]"
  }

   Or even more simply:

  asg \%n \frfiles(*.txt,&a)
  sort &a
  show array &a

   As noted elsewhere, the file lists built by \ffiles(), \frfiles(), etc,
   are now "safe" in the sense that SEND and other file-related commands
   can reference \fnextfile() without resetting the list:

  set send pathnames relative
  for \%i 1 \frfiles(*.txt) 1 {
      asg \%a \fnextfile()
      echo Sending \%a...
      send \%a
      if fail break
  }

   Copying to an array (as shown on p.398 ofUsing C-Kermit 2nd Ed)
   is no longer necessary.

4.11.5. Moving Directory Trees Between Like Systems

4.11.5.1. UNIX to UNIX

   Transferring a directory tree from one computer to another replicates
   the file sender's arrangement of files and directories on the file
   receiver's computer. Normally this is done using relative pathnames,
   since the user IDs might not be identical on the two computers. Let's
   say both computers are UNIX based, running C-Kermit 7.0 or later. On
   the sending computer (leaving out the connection details, etc):

  C-Kermit> cd /usr/olga
  C-Kermit> send /recursive .

   The /RECURSIVE switch tells C-Kermit to descend through the directory
   tree and to include relative pathnames on outbound filenames.

   On the receiving computer:

  C-Kermit> mkdir olgas-files           ; Make a new directory.
  C-Kermit> cd olgas-files              ; CD to it.
  C-Kermit> receive /recursive          ; = /PATHNAMES:RELATIVE

   Each Kermit program recognizes that the other is running under UNIX and
   switches to binary mode and literal filenames automatically.
   Directories are automatically created on the receiving system as
   needed. File dates and permissions are automatically reproduced from
   source to destination.

4.11.5.2. VMS to VMS

   To send recursively from VMS, simply include the /RECURSIVE switch, for
   example at the sender:

  $ kermit
  C-Kermit> cd [olga]
  C-Kermit> send /recursive *.*;0

   And at the receiver:

  C-Kermit> cd [.olga]
  C-Kermit> receive /recursive

   The notation "..." within directory brackets in VMS means "this
   directory and all directories below it"; the /RECURSIVE switch, when
   given to the sender, implies the use of "..." in the file specification
   so you don't have to include "..."; but it makes no difference if you
   do:

  $ kermit
  C-Kermit> send /recursive [olga...]*.*;0

   And at the receiver:

  C-Kermit> cd [.olga]
  C-Kermit> receive /recursive

   In either case, since both systems recognize each other as VMS, they
   switch into LABELED transfer mode automatically.

4.11.6. Moving Directory Trees Between Unlike Systems

   There are several difficulties with recursive transfers between unlike
   systems:

     * File formats can be different, especially text files character sets
       and record formats. This can now be handled by using SET FILE
       PATTERN, SET FILE TEXT-PATTERNS, and SET FILE BINARY-PATTERNS
       ( Section 4.3).
     * File naming conventions are different. For example, one system
       might allow (and use) longer filenames than the other. You can tell
       Kermit how to handle file names with the normal "set file names"
       and "set file collision" mechanisms. Most modern Kermits are fairly
       tolerant of illegal filenames and should not fail simply because of
       an incoming filename; rather, it will do its best to convert it to
       a recognizable and unique legal filename.
     * Directory notations can be different, e.g. backslashes instead of
       slashes, brackets, parentheses, spaces, etc. But this is now
       handled by converting pathnames to a standard format during
       transfer ( Section 4.10).

   So now, for the first time, it is possible to send directory trees
   among any combination of UNIX, DOS, Windows, OS/2, VMS, AOS/VS, etc.
   Here's an example sending files from an HP-UX system (where text files
   are encoded in the HP Roman8 character set) to a PC with K95 (where
   text files are encoded in CP850):

 Sender:
  cd xxx                           ; CD to root of source tree
  set file type binary             ; Default transfer mode
  set file character-set hp-roman8 ; Local character set for text files
  set xfer character-set latin1    ; Transfer character set
  set file patterns on             ; Enable automatic file-type switching...
  set file binary-patterns *.Z *.gz *.o  ; based on these patterns...
  set file text-patterns *.txt *.c *.h   ; for binary and text files.
  send /recursive *                ; Send all the file in this directory tree

 Receiver:
  cd yyy                           ; CD to root of destination tree
  set file character-set cp850     ; Local character set for text files
  receive /pathnames:relative      ; Receive with pathnames

   Notes:
     * Replace "xxx" and "yyy" with the desired directories.
     * Replace the file character sets appropriately.
     * Change the patterns as needed (or just use the built-in default
       lists).
     * SEND /RECURSIVE also implies /PATHNAMES:RELATIVE.
     * The file sender tells the file receiver the transfer mode of each
       file.
     * The file sender tells the file receiver the transfer character set.
     * By default, destination file dates will be the same as on the
       source.
     * Many of the settings shown might already be set by default.
     * SeeSections 4.3,4.10, and4.15 for additional
       explanation.

   If you are refreshing an existing directory on the destination
   computer, use "set file collision update" or other appropriate file
   collision option to handle filename collisions.

4.12. Where Did My File Go?

   Now that Kermit can be started by clicking on desktop icons (thus
   obscuring the concept of "current directory"), and can have a download
   directory, and can create directories for incoming files on the fly,
   etc, sometimes it is easy to lose a file after transfer. Of course, if
   you keep a transaction log:

  LOG TRANSACTIONS

   it will record the fate and final resting place of each file. But in
   case you did not keep a log, the new command:

  WHERE

   added in C-Kermit 7.0, gives you as much information as it has about
   the location of the last files transferred, including the pathname
   reported by the receiving Kermit, if any, when C-Kermit is the sender.
   This information was also added to SHOW FILE in somewhat less detail.

4.13. File Output Buffer Control

   (UNIX only). The new command SET FILE OUTPUT lets you control how
   incoming files are written to disk:

   SET FILE OUTPUT BUFFERED [ size ]
          Chooses buffered file output; this is the default. UNIX does its
          normal sort of disk buffering. The optional size specifies
          Kermit's own file output buffer size, and therefore the
          frequency of disk accesses (write() system calls) -- the bigger
          the size, the fewer the disk accesses.

   SET FILE OUTPUT UNBUFFERED [ size ]
          This forces each file output write() call to actually commit the
          data to disk immediately. Choosing this option will usually slow
          file reception down.

   SET FILE OUTPUT BLOCKING
          Write() calls should not return until they are complete. This is
          the normal setting, and it lets Kermit detect disk-write errors
          immediately.

   SET FILE OUTPUT NONBLOCKING
          Write() calls should return immediately. This can speed up file
          reception, but also delay the detection of disk-write errors.

   Experimentation with these parameters should be harmless, and might (or
   might not) have a perceptible, even dramatic, effect on performance.

4.14. Improved Responsiveness

   In version 7.0, C-Kermit's file-transfer protocol engine has been tuned
   for additional speed and responsiveness.

     * Binary-mode transfers over 8-bit connections, a very common case,
       are now handled in a special way that minimizes overhead.
     * SET TRANSFER CRC-CALCULATION is now OFF by default, rather than ON.
       (This affects only the overall per-transfer CRC, \v(crc16), not the
       per-packet CRCs)
     * Connection loss during file transfer is now detected immediately in
       most cases on Internet connections and on serial connections when
       CARRIER-WATCH is not set to OFF.

4.15. Doubling and Ignoring Characters for Transparency

   The following commands were added in 7.0, primarily to allow successful
   file transfer through ARPAnet TACs and with Honeywell DPS6 systems, but
   can be used in any setting where they might be needed:

   SET SEND DOUBLE-CHAR { [ char [ char [ ... ] ] ], NONE }
          Tells C-Kermit to double the specified characters (use decimal
          notation) in packets that it sends. For example, if you are
          sending files through a device that uses @ as an escape
          character, but allows you to send a single copy of @ through by
          doubling it, use "set send double 64".

   SET RECEIVE IGNORE-CHAR [ char [ char [ ... ] ] ]
          Tells C-Kermit to ignore the specified character(s) in incoming
          packets. Use this, for example, when something between the
          sender and receiver is inserting linefeeds for wrapping, NULs
          for padding, etc.

4.16. New File-Transfer Display Formats

   SET TRANSFER DISPLAY { BRIEF, CRT, FULLSCREEN, NONE, SERIAL }
          Selects the file-transfer display format.

   BRIEF is the new one. This writes one line to the screen per file,
   showing the file's name, transfer mode, size, the status of the
   transfer, and when the transfer is successful, the effective data rate
   in characters per second (CPS). Example:

  SEND ckcfn3.o (binary) (59216 bytes): OK (0.104 sec, 570206 cps)
  SEND ckcfns.o (binary) (114436 bytes): OK (0.148 sec, 772006 cps)
  SEND ckcmai.c (text) (79147 bytes): OK (0.180 sec, 438543 cps)
  SEND ckcmai.o (binary) (35396 bytes): OK (0.060 sec, 587494 cps)
  SEND ckcnet.o (binary) (62772 bytes): REFUSED
  SEND ckcpro.o (binary) (121448 bytes): OK (0.173 sec, 703928 cps)
  SEND ckcpro.w (text) (63687 bytes): OK (0.141 sec, 453059 cps)
  SEND makefile (text) (186636 bytes): OK (0.444 sec, 420471 cps)
  SEND wermit (binary) (1064960 bytes): OK (2.207 sec, 482477 cps)

   Note that transfer times are now obtained in fractional seconds, rather
   than whole seconds, so the CPS figures are more accurate (the display
   shows 3 decimal places, but internally the figure is generally precise
   to the microsecond).

4.17. New Transaction Log Formats

   The new command:

  SET TRANSACTION-LOG { VERBOSE, FTP, BRIEF [ separator ] }

   lets you choose the format of the transaction log. VERBOSE (the
   default) indicates the traditional format described in the book. BRIEF
   and FTP are new. This command must be given prior to the LOG
   TRANSACTION command if a non-VERBOSE type is desired.

4.17.1. The BRIEF Format

   BRIEF chooses a one-line per file format suitable for direct
   importation into databases like Informix, Oracle, or Sybase, in which:

     * Each record has 8 fields.
     * Fields are separated by a non-alphanumeric separator character.
     * The default separator character is comma (,).
     * Any field containing the separator character is enclosed in
       doublequotes.
     * The final field is enclosed in doublequotes.

   The fields are:

    1. Date in yyyymmdd format
    2. Time in hh:mm:ss format
    3. Action: SEND or RECV
    4. The local filename
    5. The size of the file
    6. The transfer mode (text, binary, image, labeled)
    7. The status of the transfer: OK or FAILED
    8. Additional status-dependent info, in doublequotes.

   Examples:

  20000208,12:08:52,RECV,/u/olga/oofa.txt,5246,text,OK,"0.284sec 18443cps"
  20000208,12:09:31,SEND,/u/olga/oofa.exe,32768,binary,OK,"1.243sec 26362cps"
  20000208,12:10:02,SEND,"/u/olga/a,b",10130,text,FAILED,"Refused: date"

   Note how the filename is enclosed in doublequotes in the final example,
   because it contains a comma.

   To obtain BRIEF format, you must give the SET TRANSACTION-LOG BRIEF
   command before the LOG TRANSACTIONS command. (If you give them in the
   opposite order, a heading is written to the log by the LOG command.)

4.17.2. The FTP Format

   SET TRANSACTION-LOG FTP (available only in UNIX) chooses a format that
   is compatible with the WU-FTPD (Washington University FTP daemon) log,
   and so can be processed by any software that processes the WU-FTPD log.
   It logs only transfers in and out, both successful and failed (but
   success or failure is not indicated, due to lack of a field in the
   WU-FTPD log format for this purpose). Non-transfer events are not
   recorded.

   Unlike other logs, the FTP-format transaction log is opened in append
   mode by default. This allows you to easily keep a record of all your
   kermit transfers, and it also allows the same log to be shared by
   multiple simultaneous Kermit processes or (permissions permitting)
   users. You can, of course, force creation of a new logfile by
   specifying the NEW keyword after the filename, e.g.

  log transactions oofa.log new

   All records in the FTP-style log are in a consistent format. The first
   field is fixed-length and contains spaces; subsequent fields are
   variable length, contain no spaces, and are separated by one or more
   spaces. The fields are:

   Timestamp
          This is an asctime-style timestamp, example: "Wed Sep 16
          20:19:05 1999" It is always exactly 24 characters long, and the
          subfields are always in fixed positions.

   Elapsed time
          The whole number of seconds required to transfer the file, as a
          string of decimal digits, e.g. "24".

   Connection
          The name of the network host to which C-Kermit is connected, or
          the name of the serial device through which it has dialed (or
          has a direct connection), or "/dev/tty" for transfers in remote
          mode.

   Bytes transferred
          The number of bytes transferred, decimal digits, e.g. "1537904".

   Filename
          The name of the file that was transferred, e.g.
          "/pub/ftp/kermit/a/README.TXT". If the filename contains any
          spaces or control characters, each such character is replaced by
          an underscore ('_') character.

   Mode
          The letter 'b' if the file was transferred in binary mode, or
          'a' if it was transferred in text (ASCII) mode.

   Options
          This field always contains an underscore ('_') character.

   Direction
          The letter 'o' if the file was transferred Out, and 'i' if the
          file was transferred In.

   User class
          The letter 'r' indicates the file was transferred by a Real
          user.

   User identification
          The ID of the user who transferred the file.

   Server identification
          The string "kermit". This distinguishes a Kermit transfer log
          record from a WU-FTPD record, which contains "ftp" in this
          field.

   Authentication class
          The digit '1' if we know the user's ID on the client system,
          otherwise '0'. Currently, always '0'.

   Authenticated user
          If the authentication class is '1', this is the user's ID on the
          client system. Otherwise it is an asterisk ('*'). Currently it
          is always an asterisk.

   Examples:

  Thu Oct 22 17:42:48 1998 0 * 94 /usr/olga/new.x a _ i r olga kermit 0 *
  Thu Oct 22 17:51:29 1998 1 * 147899 /usr/olga/test.c a _ o r olga kermit 0 *
  Thu Oct 22 17:51:44 1998 1 * 235 /usr/olga/test.o b _ i r olga kermit 0 *
  Fri Oct 23 12:10:25 1998 0 * 235 /usr/olga/x.ksc a _ o r olga kermit 0 *

   Note that an ftp-format transaction log can also be selected on the
   Kermit command line as follows:

  kermit --xferfile:filespec

   This is equivalent to:

  SET TRANSACTION-LOG FTP
  LOG TRANSACTIONS filespec APPEND

   Conceivably it could be possible to have a system-wide shared Kermit
   log, except that UNIX lacks any notion of an append-only file; thus any
   user who could append to the log could also delete it (or alter it).
   This problem could be worked around using setuid/setgid tricks, but
   these would most likely interfere with the other setuid/setgid tricks
   C-Kermit must use for getting at dialout devices and UUCP logfiles.

4.18. Unprefixing NUL

   As of 6.1.193 Alpha.10, C-Kermit can finally send and receive
   file-transfer packets in which NUL (ASCII 0) is unprefixed (no more
   NUL-terminated packets!). NUL is, of course, extremely prevalent in
   binary files such as executables, and this has been a significant
   source of packet overhead. For example, when transferring itself (the
   SunOS C-Kermit executable) with minimal prefixing and 9000-byte
   packets, we see:

  File size:                       1064960
  Packet chars with 0 prefixed:    1199629  overhead = 12.65%
  Packet chars with 0 unprefixed:  1062393  overhead = -0.03%

   Transfer rates go up accordingly, not only because of the reduced
   amount of i/o, but also because less computation is required on each
   end.

4.19. Clear-Channel Protocol

   Now that C-Kermit itself is capable of sending and receiving any byte
   at all on a clear channel ( Section 4.18), it is, for the first
   time, in a position to negotiate a clear channel with the other Kermit,
   giving it permission (but not requiring it) to unprefix any and all
   characters that it knows are safe. In general this means all but the
   Kermit start-of-packet character (normally Ctrl-A), Carriage Return
   (not only Kermit's end-of-packet character, but also treated specially
   on Telnet NVT links), and IAC (255, also special to Telnet).

   By default, C-Kermit will say it has a clear channel only if it has
   opened a TCP socket. Since the Kermit program on the far end of a
   TCP/IP connection generally does not know it has a TCP/IP connection,
   it will not announce a clear channel unless it has been told to do so.
   The command is:

   SET CLEAR-CHANNEL { ON, OFF, AUTO }

   AUTO is the default, meaning that the clear-channel status is
   determined automatically from the type of connection. ON means to
   announce a clear channel, OFF means not to announce it. Use SHOW
   STREAMING ( Section 4.20) to see the current CLEAR-CHANNEL status.
   Synonym: SET CLEARCHANNEL.

   CLEAR-CHANNEL is also set if you start C-Kermit with the -I switch (see
  Section 4.20).

   Whenever a clear channel is negotiated, the resulting control-character
   unprefixing is "sticky"; that is, it remains in effect after the
   transfer so you can use SHOW CONTROL to see what was negotiated.

   You can also see whether a clear channel was negotiated in the
   STATISTICS /VERBOSE Display.

   The advantage of the clear channel feature is that it can make file
   transfers go faster automatically. The disadvantage would be
   file-transfer failures if the channel is not truly clear, for example
   if C-Kermit made a Telnet connection to a terminal server, and then
   dialed out from there; or if C-Kermit made an Rlogin connection to host
   and then made a Telnet connection from there to another host. If a file
   transfer fails on a TCP/IP connection, use SHOW CONTROL to check
   whether control characters became unprefixed as a result of protocol
   negotiations, and/or SHOW STREAMING ( Section 4.20) to see if
   "clear-channel" was negotiated. If this happened, use SET CLEAR-CHANNEL
   OFF and SET PREFIXING CAUTIOUS (or whatever) to prevent it from
   happening again.

4.20. Streaming Protocol

   A new Kermit protocol option called "streaming" was added in C-Kermit
   7.0. The idea is that if the two Kermit partners have a reliable
   transport (such as TCP/IP or X.25) between them, then there is no need
   to send ACKs for Data packets, or NAKs, since a reliable transport
   will, by definition, deliver all packets in order and undamaged. On
   such a connection, streaming cuts down not only on Kermit program
   overhead (switching back and forth between reading and sending
   packets), but also tends to make the underlying transport use itself
   more efficiently (e.g. by defeating the Nagle algorithm and/or Delayed
   ACK stratagem of the TCP layer). Furthermore, it allows transfers to
   work smoothly on extremely slow network congestions that would
   otherwise cause timeouts and retransmissions, and even failure when the
   retry limit was exceeded.

   The trick is knowing when we can stream:

    1. If C-Kermit has opened a TCP socket or X.25 connection, it offers
       stream.
    2. If C-Kermit has been started with the -I (uppercase) option, or if
       it has been told to SET RELIABLE ON, it offers to stream.
    3. If C-Kermit is in remote mode, and has been told to SET RELIABLE
       AUTO (or ON), it always offers to stream, and also always agrees to
       stream, if the other Kermit offers. Unless you take explicit
       actions to override the defaults, this allows the local Kermit (the
       one that made the connection, and so knows whether it's reliable)
       to control streaming.

   (Note that an offer to stream also results in a Clear-Channel
   announcement if CLEAR-CHANNEL is set to AUTO; see Section 4.19.)

   When BOTH Kermits offer to stream, then they stream; otherwise they
   don't. Thus streaming-capable Kermit programs interoperate
   automatically and transparently with nonstreaming ones. If the two
   Kermits do agree to stream, you'll see the word "STREAMING" on the
   fullscreen file-transfer display in the Window Slots field. You can
   also find out afterwards with the STATISTICS or SHOW STREAMING
   commands.

     WARNING: Automatic choice of streaming is based on the assumption of
     a "direct" end-to-end network connection; for example, a Telnet or
     Rlogin connection from host A to host B, and transferring files
     between A and B. However, if your connection has additional
     components -- something "in the middle" (B) that you have made a
     network connection to, which makes a separate connection to the
     destination host (C), then you don't really have a reliable
     connection, but C-Kermit has no way of knowing this; transferring
     files between A and C will probably fail. In such cases, you'll need
     to tell the *local* C-Kermit to "set reliable off" before
     transferring files (it does no good to give this command to the
     remote Kermit since the local one controls the RELIABLE setting).

   Streaming is like using an infinite window size, with no timeouts and
   no tolerance for transmission errors (since there shouldn't be any). It
   relies on the underlying transport for flow control, error correction,
   timeouts, and retransmission. Thus it is very suitable for use on
   TCP/IP connections, especially slow or bursty ones, since Kermit's
   packet timeouts won't interfere with the transfer -- each packet takes
   as long to reach its destination as it takes TCP to deliver it. If TCP
   can't deliver the packet within its own timeout period (over which
   Kermit has no control), it signals a fatal error. Just like FTP.

   Streaming goes much faster than non-streaming when a relatively small
   packet length is used, and it tends to go faster than non-streaming
   with even the longest packet lengths. The Kermit window size is
   irrelevant to streaming protocol, but still might affect performance in
   small ways since it can result in different paths through the code.

   The definition of "reliable transport" does not necessarily demand
   8-bit and control-character transparency. Streaming can work with
   parity and/or control-character prefixing just as well (but not as
   fast) as without them; in such cases you can leave RELIABLE set to ON,
   but set CLEARCHANNEL and/or PARITY appropriately.

   Maximum performance -- comparable to and often exceeding FTP -- is
   achieved on socket-to-socket connections (in which the considerable
   overhead of the terminal driver and Telnet or Rlogin server is
   eliminated) with long packets and the new "brief" file-transfer display
   ( Section 4.16).

4.20.1. Commands for Streaming

   SET RELIABLE { ON, OFF, AUTO }
          SET RELIABLE ON tells Kermit that it has a reliable transport.
          SET RELIABLE OFF tells Kermit the transport is not reliable.
          SET RELIABLE AUTO tells Kermit that it should SET RELIABLE ON
          whenever it makes a reliable connection (e.g. TELNET or SET HOST
          on a TCP/IP or X.25 network), and when in remote mode it should
          believe the transport is reliable if the other Kermit says it is
          during Kermit protocol negotiation.

   AUTO is the default; the Kermit program that makes the connection knows
   whether it is reliable, and tells the remote Kermit.

   The RELIABLE setting has several effects, including:

     * It can affect the timeouts used during normal ACK/NAK protocol.
     * It can affect the clear-channel announcement.
     * It can affect streaming.

   If you TELNET or SET HOST somewhere, this includes an implicit SET
   RELIABLE ON command. The -I command-line option is equivalent to SET
   RELIABLE ON.

   Since SET RELIABLE ON (and -I) also implies SET CLEAR CHANNEL ON, you
   might find that in certain cases you need to tell Kermit that even
   though the connection is reliable, it doesn't have a clear channel
   after all:

  SET CLEAR-CHANNEL OFF
  SET PREFIXING CAUTIOUS ; or whatever...

   You can control streaming without affecting the other items with:

  SET STREAMING { ON, OFF, AUTO }

   AUTO is the default, meaning streaming will occur if Kermit has made a
   TCP/IP connection or if RELIABLE is ON (or it was started with the -I
   command line option). OFF means don't stream; ON means offer to stream
   no matter what.

4.20.2. Examples of Streaming

   Here we look at the use and behavior of streaming on several different
   kinds of connections, and compare its performance with non-streaming
   transfers.

4.20.2.1. Streaming on Socket-to-Socket Connections

   Here we get streaming automatically when both Kermit programs are
   capable of it, since they both make socket connections. For example, on
   the far end:

  C-Kermit> set host * 3000
  C-Kermit> server

   and on the near end:

  C-Kermit> set host foo.bar.xyz.com 3000
  (now give SEND and GET command)

   All subsequent file transfers use streaming automatically.

   Here are the results from 84 trials, run on a production network,
   disk-to-disk, in which a 1-MB binary file (the SunOS C-Kermit Sparc
   executable) was sent from a Sun Sparc-10 with SunOS 4.1.3 to an IBM
   Power Server 850 with AIX 4.1, socket-to-socket, over a 10Mbps 10BaseT
   Ethernet, using minimal control-character unprefixing, window sizes
   from 10 to 32, and packet sizes from 1450 to 9010:

                Streaming    Nonstreaming
  Max CPS         748955        683354
  Min CPS         221522        172491
  Mean CPS        646134        558680
  Median CPS      678043        595874
  Std Dev         101424        111493

   Correlations:

  CPS and window size:   -0.036
  CPS and packet length:  0.254
  CPS and streaming:      0.382

   Note that the relationship between streaming and throughput is
   significantly stronger than that between CPS and window size or packet
   length.

   Also note that this and all other performance measurements in this
   section are snapshots in time; the results could be much different at
   other times when the load on the systems and/or the network is higher
   or lower.

   In a similar socket-to-socket trial, but this time over a wide-area
   TCP/IP connection (from New York City to Logan, Utah, about 2000
   miles), the following results were obtained:

                Streaming    Nonstreaming
  Max CPS         338226        318203
  Min CPS         191659        132314
  Mean CPS        293744        259240
  Median CPS      300845        273271
  Std Dev          41914         52351

   Correlations:

  CPS and window size:    0.164
  CPS and packet length:  0.123
  CPS and streaming:      0.346

4.20.2.2. Streaming on Telnet Connections

   In this case the local copy of Kermit is told to TELNET or SET HOST,
   and so it knows it has a reliable connection and -- unless it has been
   told not to -- will offer to stream, and the other Kermit program,
   since it has STREAMING set to AUTO, agrees.

   Since we have a reliable connection, we'll also get control-character
   unprefixing automatically because of the new clear-channel protocol
   ( Section 4.19).

   Any errors that occur during streaming are fatal to the transfer. The
   message is "Transmission error on reliable link". Should this happen:

    1. Check the remote Kermit's flow control setting (SHOW
       COMMUNICATIONS). If it is NONE, change it to XON/XOFF, or vice
       versa. If it is XON/XOFF (or you just changed it to XOFF/XOFF),
       make sure the file sender is prefixing the XON and XOFF characters.
       In the most drastic case, use "set prefix all" to force prefixing
       of all control characters.
    2. The remote Telnet server might chop off the 8th bit. In that case,
       tell C-Kermit to "set parity space". Or, you might be able to force
       the Telnet to allow eight-bit data by telling C-Kermit to "set
       telopt binary request accept" -- that is, request the Telnet server
       to enter binary mode, and accept binary-mode bids from the server.
    3. The remote Telnet server might have a buffering limitation. If a
       and b don't cure the problem, tell the file receiver to "set
       receive packet-length 1000" (or other number -- use the largest one
       that works). This too, is no different from the non-streaming case
       (more about this in Section 4.20.2.3).

   And remember you can continue interrupted binary-mode transfers where
   they left off with the RESEND (= SEND /RECOVER) command.

   Here are the figures for the same 84 trials between the same Sun and
   IBM hosts as in 4.20.2.1, on the same network, but over a Telnet
   connection rather than socket-to-socket:

                  Streaming    Nonstreaming
  Max CPS         350088        322523
  Min CPS          95547        173152
  Mean CPS        321372        281830
  Median CPS      342604        291469
  Std Dev          40503         29948

   Correlations:

  CPS and window size:    0.001
  CPS and packet length:  0.152
  CPS and streaming:      0.128

   Here the effect is not as emphatic as in the socket-to-socket case, yet
   on the whole streaming tends to be beneficial.

   Additional measurements on HP-UX using C-Kermit 7.0 Beta.06:

                  Windowing     Streaming
  HP-UX 8->8      not tested       14Kcps
  HP-UX 8->9      not tested       76Kcps
  HP-UX 8->10      36Kcps          66Kcps
  HP-UX 9->9      not tested      190Kcps
  HP-UX 9->10     160Kcps         378Kcps

4.20.2.3. Streaming with Limited Packet Length

   The IRIX telnet server (at least the ones observed in IRIX 5.3 and 6.2)
   does not allow Kermit to send packets longer than 4096 bytes. Thus when
   sending from IRIX C-Kermit when it is on the remote end of a Telnet
   connection, the packet length must be 4K or less. Trials in this case
   (in which packet lengths range from 1450 to 4000) show a strong
   advantage for streaming, which would be evident in any other case where
   the packet length is restricted, and stronger the shorter the maximum
   packet length.

                  Streaming    Nonstreaming
  Max CPS         426187        366870
  Min CPS         407500        276517
  Mean CPS        415226        339168
  Median CPS      414139        343803
  Std Dev           6094         25851

   Correlations:

  CPS and window size:    0.116
  CPS and packet length:  0.241
  CPS and streaming:      0.901

4.20.2.4. Streaming on Dialup Connections

   Here "dialup" refers to a "direct" dialup connection, not a SLIP or PPP
   connection, which is only a particular kind of TCP/IP connection.

   Attempt this at your own risk, and then only if (a) you have
   error-correcting modems, and (b) the connections between the modems and
   computers are also error-free, perfectly flow-controlled, and free of
   interrupt conflicts. Streaming can be used effectively and to fairly
   good advantage on such connections, but remember that the transfer is
   fatal if even one error is detected (also remember that should a
   binary-mode transfer fail, it can be recovered from the point of
   failure with RESEND).

   To use streaming on an unreliable connection, you must tell both
   Kermits that the connection is reliable:

  kermit -I

   or:

  C-Kermit> set reliable on

   In this case, it will probably be necessary to prefix some control
   characters, for example if your connection is through a terminal server
   that has an escape character. Most Cisco terminal servers, for example,
   require Ctrl-^ (30, as well as its high-bit equivalent, 158) to be
   prefixed. To unprefix these, you'll need to defeat the "clear channel"
   feature:

  C-Kermit> set reliable on
  C-Kermit> set clear-channel off
  C-Kermit> set prefixing none
  C-Kermit> set control prefix 1 13 30 158 ; and whatever else is necessary

   Dialup trials were done using fixed large window and packet sizes. They
   compare uploading and downloading of two common types of files, with
   and without streaming. Configuration:

     HP-9000/715/33 -- 57600bps, RTS/CTS -- USR Courier V.34 --
     V.34+V.42, 31200bps -- USR V.34+ Rackmount -- 57600bps, RTS/CTS --
     Cisco terminal server -- Solaris 2.5.1. Packet size = 8000, Window
     Size = 30, Control Character Unprefixing Minimal (but including the
     Cisco escape character).

   Since this is not a truly reliable connection, a few trials failed when
   a bad packet was received (most likely due to UART overruns); the
   failure was graceful and immediate, and the message was informative.
   The results of ten successful trials uploading and downloading the two
   files with and without streaming are:

            Streaming..
            Off    On
   Upload   5194   5565   txt (= C source code, 78K)
            3135   3406   gz  (= gzip file, compressed, 85K)
 Download   5194   5565   txt
            3041   3406   gz

   Each CPS figure is the mean of 10 results.

   A brief test was also performed on a LAT-based dialout connection from
   a VAX 3100 with VMS 5.5 to a USR Courier V.34 connected to a DECserver
   700 at 19200 bps. The 1-MB Sparc executable downloaded from a Sun to
   the VAX at 1100cps without streaming and 1900cps with streaming, using
   8000-byte packets, 30 window slots, and minimal prefixing in both
   cases.

4.20.2.5. Streaming on X.25 Connections

   We have only limited access to X.25 networks. One trial was performed
   in which the 1MB Solaris 2.4 Sparc executable was transferred over a
   SunLink X.25 connection; nothing is known about the actual physical
   connection. With a packet length of 8000 and a window size of 30, the
   file transferred at 6400 cps (using a maximum of 6 window slots). With
   the same packet length, but with streaming, it transferred without
   mishap at 6710 cps, about 5% faster.

4.20.3. Streaming - Preliminary Conclusions

   The results vary with the particular connection, but are good overall.
   Although numerous lower-level tricks can be used to improve performance
   on specific platforms or connection methods, streaming occurs at a
   high, system-independent level of the Kermit protocol and therefore can
   apply to all types of platforms and (reliable) connections
   transparently.

4.21. The TRANSMIT Command

   Prior to C-Kermit 7.0, the TRANSMIT command transmitted in text or
   binary mode according to SET FILE TYPE { TEXT, BINARY }. But now that
   binary mode is likely to be the default for protocol transfers, it is
   evident that this not also an appropriate default for TRANSMIT, since
   binary-mode TRANSMIT is a rather specialized and tricky operation.
   Therefore, TRANSMIT defaults to text mode always, regardless of the
   FILE TYPE setting.

   C-Kermit 7.0 expands the capabilities of the TRANSMIT command by adding
   the following switches (see Section 1.5). The new syntax is:

  TRANSMIT [ switches... ] filename

   Zero or more switches may be included:

   /PIPE
          When /PIPE is included, "filename" is interpreted as a system
          command or program whose output is to be sent. Synonym:
          /COMMAND. Example:

  transmit /pipe finger

          You may enclose the command in braces, but you don't have to:

  xmit /pipe {ls -l | sort -r +0.22 -0.32 | head}

   /BINARY
          Transmits the file (or pipe output) in binary mode.

   /TEXT
          Transmits the file (or pipe output) in line-oriented text mode.
          Current FILE CHARACTER-SET and TERMINAL CHARACTER-SET selections
          govern translation. Default.

   /TRANSPARENT
          Specifies text mode without character-set translation, no matter
          what the FILE and TERMINAL CHARACTER-SET selections are.

   /NOWAIT
          This is equivalent to SET TRANSMIT PROMPT 0, but for this
          TRANSMIT command only. Applies only to text mode; it means to
          not wait for any kind of echo or turnaround character after
          sending a line before sending the next line. (Normally Kermit
          waits for a linefeed.)

   When TRANSMIT ECHO is ON, C-Kermit tries to read back the echo of each
   character that is sent. Prior to C-Kermit 7.0, 1 second was allowed for
   each echo to appear; if it didn't show up in a second, the TRANSMIT
   command would fail. Similarly for the TRANSMIT PROMPT character.
   However, with today's congested Internet connections, etc, more time is
   often needed:

   SET TRANSMIT TIMEOUT number

   Specifies the number of seconds to wait for an echo or the prompt
   character when TRANSMIT PROMPT is nonzero; the default wait is 1
   second. If you specify 0, the wait is indefinite. When a timeout
   interval of 0 is specified, and a desired echo or prompt does not show
   up, the TRANSMIT command will not terminate until or unless you
   interrupt it with Ctrl-C; use SET TRANSMIT TIMEOUT 0 with caution.

   Note: to blast a file out the communications connection without any
   kind of synchronization or timeouts or other manner of checking, use:

  SET TRANSMIT ECHO OFF
  SET TRANSMIT PROMPT 0 (or include the /NOWAIT switch)
  SET TRANSMIT PAUSE 0
  TRANSMIT [ switches ] filename

   In this case, text-file transmission is not-line oriented and large
   blocks can be sent, resulting in a significant performance improvement
   over line-at-at-time transmission. Successful operation depends (even
   more than usual for the TRANSMIT command!) on a clean connection with
   effective flow control.

   For details on TRANSMIT and character sets, see Section 6.6.5.4.

  4.22. Coping with Faulty Kermit Implementations

   Kermit protocol has been implemented in quite a few third-party
   commercial, shareware, and freeware software packages, with varying
   degrees of success. In most cases operation is satisfactory but slow --
   only the bare minimum subset of the protocol is available -- short
   packets, no sliding windows, no attributes, etc. In other cases, the
   implementation is incorrect, resulting in failures at the initial
   negotiation stage or corrupted files.

   C-Kermit 7.0 and Kermit 95 1.1.19 include some new defense mechanisms
   to help cope with the most common situations. However, bear in mind
   there is only so much we can do in such cases -- the responsibility for
   fixing the problem lies with the maker of the faulty software.

    4.22.1. Failure to Accept Modern Negotiation Strings

   The published Kermit protocol specification states that new fields can
   be added to the parameter negotiation string. These are to be ignored
   by any Kermit implementation that does not understand them; this is
   what makes the Kermit protocol extensible. Unfortunately, some Kermit
   implementations become confused (or worse) when receiving a negotiation
   string longer than the one they expect. You can try working around such
   problems by telling Kermit to shorten its negotiation string (and thus
   disable the corresponding new features):

  SET SEND NEGOTIATION-STRING-MAX-LENGTH number

   Try a number like 10. If that doesn't work, try 9, 8, 7, 6, and so on.

    4.22.2. Failure to Negotiate 8th-bit Prefixing

   The published Kermit protocol specification states that 8th-bit
   prefixing (which allows transfer of 8-bit data over a 7-bit connection)
   occurs if the file sender puts a valid prefix character (normally "&")
   in the 8th-bit-prefix field of the negotiation string, and the receiver
   puts either a letter "Y" or the same prefix character. At least one
   faulty Kermit implementation exists that does not accept the letter
   "Y". To force C-Kermit / K-95 to reply with the other Kermit's prefix
   character rather than a "Y", give the following (invisible) command:

  SET Q8FLAG ON

   Use SET Q8FLAG OFF to restore the normal behavior.

    4.22.3. Corrupt Files

   Refer to Section 4.22.2. Some Kermit implementations mistakenly
   interpret the "Y" as a prefix character. Then, whenever a letter Y
   appears in the data, the Y and the character that follows it are
   replaced by a garbage character. At this writing, we are not sure if
   there is any solution, but try "set send negotiation-string-max-length
   6" and/or "set q8flag on".

   File corruption can also occur when control characters within the file
   data are sent without prefixing, as at least some are by default in
   C-Kermit 7.0 and K-95. Some Kermit implementations do not handle
   incoming "bare" control characters. To work around, "set prefixing
   all".

    4.22.4. Spurious Cancellations

   The Kermit protocol specification states that if an ACK to a Data
   packet contains X in its data field, the transfer of the current file
   is canceled, and if it contains a Z, the entire transfer is canceled.
   At least one overzealous Kermit implementation applies this rule to
   non-Data packets as well, the typical symptom being that any attempt to
   transfer a file whose name begins with X or Z results in cancellation.
   This is because the file receiver typically sends back the name under
   which it stored the file (which might not be the same as the name it
   was sent with) in the ACK to the File Header packet. This is
   information only and should not cause cancellation. To work around the
   problem, use:

  SET F-ACK-BUG { ON, OFF }

   ON tells Kermit not to send back the filename in the ACK to the file
   header packet as it normally would do (OFF puts Kermit back to normal
   after using ON).

   A variation on the this bug occurs in an obscure Kermit program for
   MUMPS: When this Kermit program sends a file called (say) FOO.BAR, it
   requires that the ACK to its F packet contain exactly the same name,
   FOO.BAR. However, C-Kermit likes to send back the full pathname,
   causing the MUMPS Kermit to fail. SET F-ACK-BUG ON doesn't help here.
   So a separate command has been added to handle this situation:

  SET F-ACK-PATH { ON, OFF }

   Normally it is ON (regardless of the SET SEND PATHNAMES setting). Use
   SET F-ACK-PATH OFF to instruct Kermit to send back only the filename
   without the path in the ACK to the F packet.

    4.22.5. Spurious Refusals

   Some Kermit implementations, notably PDP-11 Kermit 3.60 and earlier,
   have bugs in their handling of Attribute packets that can cause
   unwarranted refusal of incoming files, e.g. based on date or size. This
   can be worked around by telling one or both of the Kermit partners to:

  SET ATTRIBUTES OFF

    4.22.6. Failures during the Data Transfer Phase

   This can be caused by control-character unprefixing (Section
   4.22.3 ), and fixed by:

  SET PREFIXING ALL

   It can also have numerous other causes, explained in Chapter 10 of
  Using C-Kermit: the connection is not 8-bit transparent (so use
   "set parity space" or somesuch), inadequate flow control, etc. Consult
   the manual.

    4.22.7. Fractured Filenames

   At least one well-known PC-based communications package negotiates data
   compression, which (according to the protocol specification) applies to
   both the filename and the file data, but then fails to decompress the
   filename. Example: C-Kermit sends a file called R000101.DAT (where
   000101 might be non-Y2K-wise YYMMDD notation), and the package in
   question stores the files as R~#0101.DAT. Workaround: Tell C-Kermit to
   SET REPEAT COUNTS OFF.

    4.22.8. Bad File Dates

   At least one well-known PC-based communications package negotiates the
   passing of file timestamps from sender to receiver, but when it is
   sending files, it always gives them a timestamp of 1 February 1970.
   Workaround: tell C-Kermit to SET ATTRIBUTE DATE OFF. You don't get the
   file's real date, but you also don't get 1 Feb 1970; instead the file
   gets the current date and time.

  4.23. File Transfer Recovery

   Prior to C-Kermit 7.0, RESEND (SEND /RECOVER) and REGET (GET /RECOVER)
   refused to work if FILE TYPE was not BINARY or the /BINARY switch was
   not included. Now these commands include an implied /BINARY switch,
   meaning they set the file type to binary for the duration of the
   command automatically.

   In the client/server arrangement, this also forces the server into
   binary mode (if it is C-Kermit 7.0 or greater, or K95 1.1.19 or
   greater) so the recovery operation proceeds, just as you asked and
   expected.

   BUT... Just as before, the results are correct only under the following
   conditions:

     * If the prior interrupted transfer was also in binary mode; or:
     * If the prior transfer was in text mode and the other computer was a
       "like platform" (e.g. UNIX-to-UNIX, Windows-to-Windows,
       DOS-to-Windows) AND there was no character-set translation (i.e.
       TRANSFER CHARACTER-SET was TRANSPARENT).

   Note that these circumstances are more likely to obtain in C-Kermit
   7.0, in which:

     * The default FILE TYPE in C-Kermit 7.0 is BINARY.
     * The default FILE INCOMPLETE setting is AUTO, which means KEEP if
       the transfer is in binary mode, DISCARD otherwise.
     * C-Kermit 7.0, Kermit 95 1.1.17, and MS-DOS Kermit 3.15 and later
       can recognize "like platforms" and switch into binary mode
       automatically. Transfers between like platforms are always binary
       unless character-set translation has been requested, and then is
       still binary for all files whose names match a binary pattern,
       unless the automatic mechanisms have been disabled (with a /TEXT
       switch, or with SET TRANSFER MODE MANUAL).
     * SEND /BINARY and GET /BINARY always force binary-mode transfers,
       even when FILE TYPE is TEXT, even when TRANSFER MODE is AUTOMATIC,
       even when PATTERNS are ON and the file's name matches a text
       pattern.

   But also note that the automatic client/server transfer-mode
   adjustments do not work with versions of C-Kermit prior to 7.0 or K95
   prior to 1.1.16.

   If the prior transfer was in text mode:

     * If text-mode transfers between the two platforms are
       "length-changing" (as they are between UNIX -- which terminates
       text lines with LF -- and DOS or Windows -- which terminates text
       lines with CRLF), the recovered file will be corrupt.
     * If text-mode transfers between the two platforms are not
       length-changing, but character-set translation was active in the
       prior transfer, the result will be a file in which the first part
       has translated characters and the second part does not.

   But in C-Kermit 7.0 and K95 1.1.19 and later, incompletely transferred
   text files are not kept unless you change the default. But if you have
   done this, and you have an incompletely transferred text file, you'll
   need to:

     * Transfer the whole file again in text mode, or:
     * Use SEND /STARTING-AT: to recover the transfer at the correct
       point; but you have to find out what that point is, as described in
       the manual.

   Kermit has no way of knowing whether the previous transfer was in text
   or binary mode so it is your responsibility to choose the appropriate
   recovery method.

   If you use C-Kermit to maintain parallel directories on different
   computers, using SET FILE COLLISION to transfer only those files that
   changed since last time, and the files are big enough (or the
   connection slow enough) to require SEND /RECOVER to resume interrupted
   transfers, you should remember that SEND /RECOVER (RESEND) overrides
   all FILE COLLISION settings. Therefore you should use SEND /RECOVER
   (RESEND) only on the file that was interrupted, not the file group. For
   example, if the original transfer was initiated with:

  SEND *

   and was interrupted, then after reestablishing your connection and
   starting the Kermit receiver with SET FILE COLLISION UPDATE on the
   remote end, use the following sequence at the sender to resume the
   transfer:

  SEND /RECOVER name-of-interrupted-file

   and then:

  SEND *

   (In C-Kermit 7.0 and later, \v(filename) contains the name of the file
   most recently transferred, as long you have not EXITed from Kermit or
   changed directory, etc.

  4.24. FILE COLLISION UPDATE Clarification

   In UNIX, file modification dates are used when comparing the file date
   with the date in the attribute packet. In VMS, however, the file
   creation date is used. These two policies reflect the preferences of
   the two user communities.

   Also, remember that the file date/time given in the attribute packet is
   the local time at the file sender. At present, no timezone conversions
   are defined in or performed by the Kermit protocol. This is primarily
   because this feature was designed at a time when many of the systems
   where Kermit runs had no concept of timezone, and therefore would be
   unable to convert (say, to/from GMT or UTC or Zulu time).

   As a consequence, some unexpected results might occur when transferring
   files across timezones; e.g. commands on the target system that are
   sensitive to file dates might not work (UNIX "make", backups, etc).

   Timezone handling is deferred for a future release.

  4.25. Autodownload Improvements

   Refer to pages 164-165 ofUsing C-Kermit about the hazards of
   autodownload when C-Kermit is "in the middle". As of C-Kermit 7.0, no
   more hazards. If C-Kermit has TERMINAL AUTODOWNLOAD ON and it detects a
   packet of the current protocol type (Kermit or Zmodem), it "erases" the
   visual aspect of the packet that would be seen by the terminal (or,
   more to the point, the emulator, such as K95). This way, only C-Kermit
   goes into RECEIVE mode, and not also the terminal emulator through
   which C-Kermit is accessed. And therefore, it is no longer necessary to
   SET TERMINAL AUTODOWNLOAD OFF to prevent multiple Kermits from going
   into receive mode at once, but of course it is still necessary to
   ensure that, when you have multiple Kermits in a chain, that the
   desired one receives the autodownload.

   The defaults have not been changed; Kermit 95 still has autodownload ON
   by default, and C-Kermit has it OFF by default.

  5. CLIENT/SERVER

  5.0. Hints

   If you use SET SERVER GET-PATH to set up your server, and the GET-PATH
   does not include the server's current directory, clients can become
   quite confused. For example, "remote dir oofa.txt" shows a file named
   oofa.txt, but "get oofa.txt" fails. In this situation, you should
   either DISABLE DIR or make your GET-PATH include the current directory.

  5.1. New Command-Line Options

   The -G command-line option is like -g (GET), except the incoming file
   is sent to standard output rather than written to disk.

   The -I option ("Internet") is used to tell a remote C-Kermit program
   that you are coming in via Internet Telnet or Rlogin and therefore have
   a reliable connection. The -I option is equivalent to SET RELIABLE ON
   and SET FLOW NONE.

   The -O option ("Only One") tells C-Kermit to enter server mode but then
   exit after the first client operation.

   See Section 9.3 for details.

  5.2. New Client Commands

   BYE and FINISH no longer try to do anything if a connection is not
   active. Thus a sequence like "hangup" followed by "bye" or "finish"
   will no longer get stuck in a long timeout-and-retransmission cycle,
   nor will it try to open a new connection.

   REMOTE EXIT
          Similar to FINISH, except it ensures that the Kermit server
          program exits back to the operating system or shell prompt.
          (FINISH would return it to its interactive prompt if it was
          started in interactive mode, and would cause it to exit if it
          entered server mode via command-line option.) When C-Kermit is
          to be the server, you can use { ENABLE, DISABLE } EXIT to
          control the client's access to this feature.

   REMOTE MKDIR directory-name
          Tells the client to ask the server to create a directory with
          the given name, which can be absolute or relative. The syntax of
          the directory name depends on the Kermit server (seenext
          section); in all cases, it can be in the syntax of the system
          where the server is running (UNIX, VMS, DOS, etc) but newer
          servers also accept UNIX syntax, no matter what the underlying
          platform. The server will not execute this command if (a) it
          does not understand it, (b) a DISABLE MKDIR command has been
          given, or (c) a DISABLE CWD command has been given; otherwise,
          the command is executed, but will fail if the directory can not
          be created, in which cases most servers will attempt to return a
          message giving the reason for failure. The REMOTE MKDIR command
          succeeds if the remote directory is created, or if it already
          exists and therefore does not need to be created, and fails
          otherwise.

   REMOTE RMDIR directory-name
          Tells the client to ask the server to remove (delete) a
          directory with the given name. The same considerations apply as
          for REMOTE MKDIR.

   REMOTE SET FILE INCOMPLETE { DISCARD, KEEP, AUTO }
          Previously this was only available in its earlier form, REMOTE
          SET INCOMPLETE (no FILE). The earlier form is still available,
          but invisible. Also, AUTO was added, meaning KEEP if in binary
          mode, DISCARD otherwise.

   REMOTE SET TRANSFER MODE { AUTOMATIC, MANUAL }
          Tells the client to ask the server to set the given
          file-transfer mode. Automatic means (roughly): if the client and
          the server are running on the same kind of computer (e.g. both
          are on UNIX), then use binary mode automatically; if the system
          types are different, use some other method to automatically
          determine text or binary mode, such as filename pattern
          matching. MANUAL means, in this context, obey the client's FILE
          TYPE setting (TEXT or BINARY). Synonym: REMOTE SET XFER MODE.

   [ REMOTE ] QUERY KERMIT function(args...)
          Prior to C-Kermit 7.0, the arguments were not evaluated locally.
          Thus it was not possible to have the server run the function
          with client-side variables as arguments. Now:

  define \%a oofa.*
  remote query kermit files(\%a)    ; Client's \%a
  remote query kermit files(\\%a)   ; Server's \%a

   [ REMOTE ] LOGIN [ user [ password ] ]
          LOGIN is now a synonym for REMOTE LOGIN.

   LOGOUT
          This command, when given in local mode, is equivalent to REMOTE
          LOGOUT. When given at the IKSD prompt, it logs out the IKSD.
          When given at the C-Kermit prompt when it has no connection, it
          does nothing.

   Note that in C-Kermit 7.0, the REMOTE (or R) prefix is not required for
   QUERY, since there is no local QUERY command. The new top-level QUERY
   command does exactly what REMOTE QUERY (RQUERY) does.

   All REMOTE commands now have single-word shortcuts:

 Shortcut   Full Form
  RASG       REMOTE ASSIGN
  RCD        REMOTE CD
  RCOPY      REMOTE COPY
  RDEL       REMOTE DELETE
  RDIR       REMOTE DIRECTORY
  REXIT      REMOTE EXIT
  RHELP      REMOTE HELP
  RHOST      REMOTE HOST
  RPWD       REMOTE PWD
  RSET       REMOTE SET
  etc.

   The R prefix is not applied to LOGIN because there is already an RLOGIN
   command with a different meaning. It is not applied to LOGOUT either,
   since LOGOUT knows what to do in each case, and for symmetry with
   LOGIN.

    5.2.1. Remote Procedure Definitions and Calls

   This is nothing new, but it might not be obvious... REMOTE ASSIGN and
   REMOTE QUERY may be used to achieve remote procedure execution. The
   remote procedure can be defined locally or remotely.

   A remote procedure call is accomplished as noted in the previous
   section:

  [ remote ] query kermit function-name(args...)

   This invokes any function that is built in to the Kermit server, e.g.:

  [ remote ] query kermit size(foo.bar)

   returns the size of the remote file, foo.bar.

   Now note that C-Kermit includes an \fexecute() function, allowing it to
   execute any macro as if it were a built-in function. So suppose MYMACRO
   is the name of a macro defined in the server. You can execute it from
   the client as follows (the redundant "remote" prefix is omitted in the
   remaining examples):

  query kermit execute(mymacro arg1 arg2...)

   The return value, if any, is the value of the RETURN command that
   terminated execution of the macro, for example:

  define addtwonumbers return \feval(\%1+\%2)

   The client invocation would be:

  query kermit execute(addtwonumbers 3 4)
  7

   The result ("7" in this case) is also assigned to the client's
   \v(query) variable.

   To execute a remote system command or command procedure (shell script,
   etc) use:

  query kermit command(name args...)

   Finally, suppose you want the client to send a macro to the server to
   be executed on the server end. This is done as follows:

  remote assign macroname definition
  query kermit execute(macroname arg1 arg2...)

   Quoting is required if the definition contains formal parameters.

  5.3. New Server Capabilities

    5.3.1. Creating and Removing Directories

   The C-Kermit 7.0 server responds to REMOTE MKDIR and REMOTE RMDIR
   commands. The directory name may be in either the native format of the
   server's computer, or in UNIX format. For example, a server running on
   VMS with a current directory of [IVAN] can accept commands from the
   client like:

  remote mkdir olga         ; Makes [IVAN.OLGA] (nonspecific format)
  remote mkdir .olga        ; Makes [IVAN.OLGA] (VMS format without brackets)
  remote mkdir olga/        ; Makes [IVAN.OLGA] (UNIX relative format)
  remote mkdir /ivan/olga   ; Makes [IVAN.OLGA] (UNIX absolute format)
  remote mkdir [ivan.olga]  ; Makes [IVAN.OLGA] (VMS absolute format)
  remote mkdir [.olga]      ; Makes [IVAN.OLGA] (VMS relative format)

    5.3.1.1. Creating Directories

   If a directory name is given that contains more than one segment that
   does not exist, the server attempts to create all the segments. For
   example, if the client says:

  REMOTE MKDIR letters/angry

   a "letters" subdirectory is created in the server's current directory
   if it does not already exist, and then an "angry" subdirectory is
   created beneath it, if it does not already have one. This can repeated
   to any reasonable depth:

  REMOTE MKDIR a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/z/y/z

    5.3.1.2. Removing Directories

   When attempting to execute a REMOTE RMDIR, the server can remove only a
   single directory, not an entire sequence or tree. The system service
   that is called to remove the directory generally requires not only that
   the server process has write delete access, but also that the directory
   contain no files.

   In the future, a REMOTE RMDIR /RECURSIVE command (and the accompanying
   protocol) might be added. For now, use the equivalent REMOTE HOST
   command(s), if any.

    5.3.2. Directory Listings

   Directory listings are generated by C-Kermit itself, rather than by
   running the underlying system's directory command. Some control over
   the listing format can be obtained with the SET OPTIONS DIRECTORY
   command ( Section 4.5.1). The following options affect listings
   sent by the server: /[NO]HEADING, /[NO]DOTFILES, and /[NO]BACKUP. In
   UNIX and VMS, the listing is always sorted by filename. There is, at
   present, no protocol defined for the client to request listing options
   of the server; this might be added in the future.

   The server's directory listings are in the following format:

   Protection or permissions:
          In UNIX and OS-9, this is a 10-character field, left adjusted.
          In VMS it is a 22-character field, left-adjusted. In each case,
          the protection / permission codes are shown in the server
          platform's native format. In other operating systems, this field
          is not shown.

   Size in bytes:
          This is always a 10-character field. The file's size is shown as
          a decimal number, right adjusted in the field. If the file is a
          directory and its size can not be obtained, the size is shown as
          "<DIR>". Two blanks follow this field.

   Date:
          Always in yyyy-mm-dd hh:mm:ss numeric format, and therefore 19
          characters long. If the file's date/time can't be obtained,
          zeros (0) are shown for all the digits. This field is followed
          by two blanks.

   Filename:
          This field extends to the end of the line. Filenames are shown
          relative to the server's current directory. In UNIX, symbolic
          links are shown as they are in an "ls -l" listing as "linkname
          -> filename".

   In UNIX and VMS, listings are returned by the server in alphabetical
   order of filename. There are presently no other sort or selection
   options.

   However, since these are fixed-field listings, all fields can be used
   as sort keys by external sort programs. Note, in particular, that the
   format used for the date allows a normal lexical on that field to
   achieve the date ordering. For example, let's assume we have a UNIX
   client and a UNIX server. In this case, the server's listing has the
   date in columns 22-40, and thus could be sorted by the UNIX sort
   program using "sort +0.22 -0.40" or in reverse order by "sort +0.22
   -0.40r".

   Since the UNIX client can pipe responses to REMOTE commands through
   filters, any desired sorting can be accomplished this way, for example:

C-Kermit> remote directory | sort +0.22 -0.40

   You can also sort by size:

  C-Kermit> remote directory | sort +0.11 -0.19

   You can use sort options to select reverse or ascending order. "man
   sort" (in UNIX) for more information. And of course, you can pipe these
   listings through any other filter of your choice, such as grep to skip
   unwanted lines.

  5.4. Syntax for Remote Filenames with Embedded Spaces

   C-Kermit and K95, when in server mode, assume that any spaces in the
   file specification in an incoming GET command are filename separators.
   Thus if the client gives a command like:

  get {oofa.txt oofa.bin}

   or, equivalently:

  mget oofa.txt oofa.bin

   the server tries to send the two files, oofa.txt and oofa.bin. But what
   if you want the server to send you a file named, say:

  D:\HP OfficeJet 500\Images\My Pretty Picture Dot PCL

   How does the server know this is supposed to be one file and not seven?
   In this case, you need to the send file name to the server enclosed in
   either curly braces:

  {D:\HP OfficeJet 500\Images\My Pretty Picture Dot PCL}

   or ASCII doublequotes:

  "D:\HP OfficeJet 500\Images\My Pretty Picture Dot PCL"

   The method for doing this depends on your client. If your client is
   C-Kermit 7.0, any recent version of Kermit 95, or MS-DOS Kermit 3.16,
   then you have to enclose the name in braces just so the client can
   parse it, so to send braces or doublequotes to the server, you must put
   them inside the first, outside pair of braces. And you also need to
   double the backslashes to prevent them from being interpreted:

  get {{D:\\HP OfficeJet 500\\Images\\My Pretty Picture Dot PCL}}
  get {"D:\\HP OfficeJet 500\\Images\\My Pretty Picture Dot PCL"}

   To get around the requirement to double backslashes in literal
   filenames, of course you can also use:

  set command quoting off
  get {{D:\HP OfficeJet 500\Images\My Pretty Picture Dot PCL}}
  get {"D:\HP OfficeJet 500\Images\My Pretty Picture Dot PCL"}
  set command quoting on

   If you are giving a "kermit" command to the UNIX shell, you have to
   observe the shell's quoting rules, something like this:

  kermit -ig "{D:\HP OfficeJet 500\Images\My Pretty Picture Dot PCL}"

   Here, the quotes go on the outside so UNIX will pass the entire
   filename, spaces, braces, and all, as a single argument to Kermit, and
   the backslashes are not doubled because (a) the UNIX shell ignores them
   since they are in a quoted string, and (b) Kermit ignores them since
   the interactive command parser is not activated in this case.

  5.5. Automatic Orientation Messages upon Directory Change

   C-Kermit 7.0, when acting as a server, can send an orientation message
   to the client whenever the server directory changes. For example, when
   the client gives a REMOTE CD command, the server sends the contents of
   the new directory's "Read Me" file to the client's screen. The
   following commands govern this feature:

   SET SERVER CD-MESSAGE FILE name
          Given to the servr, allows the message-file name to be specified
          at runtime. A list of names to look for can be given in the
          following format:

  {{name1}{name2}{name3}{...}}

          e.g. SET SERVER CD-MESSAGE FILE
          {{./.readme}{README.TXT}{READ.ME}}

   REMOTE SET SERVER CD-MESSAGE { ON, OFF }
          Given to the client, lets the client control whether the server
          sends automatic CD messages.

   SHOW SERVER
          Given to server, includes CD-Message status.

   The default CD message file name is system dependent. SHOW CD or SHOW
   SERVER displays the list. Also see Section 4.5.2.

  5.6. New Server Controls

   DISABLE ENABLE
          Allows the server to configured such that DISABLEd features can
          not be re-enabled by any means -- e.g. if the client is somehow
          able to get the server into command mode. Once DISABLEd, ENABLE
          can not be re-ENABLEd.

   SET SERVER IDLE-TIMEOUT seconds
          This was available previously in Kermit 95 only. Now it can be
          used in C-Kermit also to specify a maximum number of seconds the
          server is allowed to be idle before exiting server mode. 0
          seconds means no idle timeout. In C-Kermit (but not K-95), SET
          SERVER TIMEOUT and SET SERVER IDLE-TIMEOUT are mutually
          exclusive -- you can have one or the other (or neither), but not
          both. (Server timeouts are for the benefit of primitive Kermit
          clients that are not capable of timing out on their own; to our
          knowledge, no such clients are still in circulation.)

   SET SERVER KEEPALIVE { ON, OFF }
          (See next section).

  5.7. Timeouts during REMOTE HOST Command Execution

   Prior to C-Kermit 7.0, the C-Kermit server would block waiting for
   output from a system command invoked via REMOTE HOST from the client.
   If the system command took a long time to execute, the client would
   time out and send NAK packets. If the command took too long, the client
   would reach its retry limit and give up. Even if it didn't, the NAKs
   would cause unnecessary retransmissions.

   In version 7.0, the C-Kermit server (VMS and select()-capable UNIX
   versions only), sends "keepalive packets" (empty data packets) once per
   second while waiting for the system command to complete. This procedure
   should be entirely transparent to the Kermit client, and should prevent
   the unwanted timeouts and NAKs. When C-Kermit 7.0 itself (or K95
   1.1.19) is the client, it prints dots to show the keepalive packets.

   The keepalive feature can be turned off and on with:

  SET SERVER KEEPALIVE { ON, OFF }

   Normally it should be on. Turn it off it if causes trouble with the
   client, or if it seems to slow down the server (as it might on some
   platforms under certain circumstances).

  6. INTERNATIONAL CHARACTER SETS

   Support for several new single-byte character sets was added in
   C-Kermit 7.0. Unicode / ISO 10646 is not yet supported, but is a high
   priority for forthcoming releases.

  6.0. ISO 8859-15 Latin Alphabet 9

   To accommodate the Euro currency symbol, and to correct several other
   longstanding problems with ISO Latin Alphabet 1, ISO 8859-15 Latin
   Alphabet 9 was issued in May 1998. It is supported by C-Kermit 7.0 as a
   transfer character set, a file character set, and a terminal character
   set. Translations that preserve the new characters are available
   between Latin-9 and several other sets including:

  PC Code Page 858         (Western European languages, similar to CP850)
  Windows Code Page 1252   (Western European languages, similar to Latin-1)
  Windows Code Page 1250   (Eastern European languages, similar to Latin-2)

   The Latin-9 transfer character set also allows for the OE digraph
   character, used primarily in French, to be preserved in transfers
   involving the DEC MCS or NeXT character sets.

   The Euro character is also present in the Universal Character Set,
   described in Section 6.6.

  6.1. The HP-Roman8 Character Set

   The HP-Roman8 character set is supported in C-Kermit 6.0 and later but
   was omitted from Table VII-4 in the 2nd Edition of Using C-Kermit due
   to lack of space. It is listed inAppendix III.

  6.2. Greek Character Sets

   Greek character sets were added in 6.1:

  SET FILE CHARACTER-SET { CP869, ELOT927, GREEK-ISO }
  SET TRANSFER CHARACTER-SET { GREEK-ISO }

   GREEK-ISO is ISO 8859-7, which the same as ELOT 928.

   The new Greek character sets are listed inAppendix III.

  6.3. Additional Latin-2 Character Sets

   The following have been added as FILE and TERMINAL CHARACTER-SETs:

   MAZOVIA-PC
          A PC code page used in Poland, equivalent to CP437, but with 18
          substitutions needed for Polish.

   CP1250
          The Windows Latin 2 Code Page. Equivalent to ISO 8859-2, but
          with different encoding.

  6.4. Additional Cyrillic Character Sets

   The following have been added as FILE and TERMINAL CHARACTER-SETs:

   BULGARIA-PC
          This is the Cyrillic PC code page used in Bulgaria, where it is
          called Code Page 856. It is attributed to a company called
          DATEC, Inc, but CP856 is not a proper designation, since it
          refers to a Hebrew Code Page (see the IBM Registry).

   CP855
          This PC Code Page contains all the Cyrillic letters that are
          also in ISO 8859-5, and is therefore useful for non-Russian
          Cyrillic text (Ukrainian, Belarusian, etc), unlike CP866, which
          has a smaller repertoire of Cyrillic letters.

   CP1251
          The Windows Cyrillic Code Page. Equivalent to CP855, but with
          different encoding.

   KOI8R
          An extension to "Old KOI-8" that adds upper and lower case
          Cyrillic letter Io (looks like Roman E with diaeresis) plus a
          selection of box-drawing characters to columns 8 through 11,
          which are vacant in original Old KOI-8. KOI8-R is used for the
          Russian language. It is specified inRFC 1489.

   KOI8U
          A similar extension of Old KOI-8, but for Ukrainian. It is
          specified inRFC 2319.

  6.5. Automatic Character-Set Switching

   Prior to version 7.0, C-Kermit's file character-set always had to be
   set explicitly. In 7.0 and later, it is set automatically when:

    1. This feature is enabled (as it is unless you disable it).
    2. An incoming text-mode transfer includes a transfer-character-set
       announcer and you have not previously given a SET FILE
       CHARACTER-SET command. In this case, C-Kermit switches to an
       appropriate file character set. For example, on an HP-UX
       workstation, an incoming Latin-1 file automatically selects
       HP-Roman8 for the local copy of the file; in Data General AOS/VS,
       it would select DG International.
    3. You give a SET TRANSFER CHARACTER-SET command without having
       previously specified a FILE CHARACTER-SET. An appropriate file
       character-set is chosen automatically.

   In addition, when you give a SET FILE CHARACTER-SET command, the
   appropriate transfer character-set is automatically chosen, to be used
   when you are sending files (but this does not override the one
   announced by the sender when you are receiving files).

   You might not agree about what is "appropriate", so of course you can
   disable or change all of the above actions.

   You can disable (or re-enable) the new automatic character-set
   switching feature in each direction separately:

   SET RECEIVE CHARACTER-SET-SELECTION { AUTOMATIC, MANUAL }
          AUTOMATIC is the default, causing the behavior described above
          when an incoming file arrives. Choose MANUAL to defeat this
          behavior and force your current FILE CHARACTER-SET setting to be
          used, no matter what it is. Note that SET RECEIVE CHARACTER-SET
          MANUAL does not disable recognition of the incoming transfer
          character-set announcer, and translation from the corresponding
          character-set to your current file character-set. To disable
          that, use SET ATTRIBUTE CHARACTER-SET OFF.

   SET SEND CHARACTER-SET-SELECTION { AUTOMATIC, MANUAL }
          Again AUTOMATIC is the default, causing the behavior described
          above when you give a SET { FILE, TRANSFER } CHARACTER-SET
          command. Use MANUAL to allow you to specify the transfer and
          file character-sets independently.

   SHOW CHARACTER-SETS
          Tells settings of { SEND, RECEIVE } CHARACTER-SET-SELECTION.

   Normally, however, it is more convenient to leave automatic switching
   active, and change any associations that are not appropriate for your
   application, area, or country. The commands are:

   SHOW ASSOCIATIONS
          This command lists all the associations in each direction: for
          each possible transfer character-set, it lists the associated
          file character-set, and vice versa. These are two separate and
          independent lists.

   ASSOCIATE TRANSFER-CHARACTER-SET name1 [ name2 ]
          Changes the association for the transfer character-set name1 to
          be the file character-set name2. If name2 is omitted, automatic
          switching is disabled for this transfer character-set only.

   ASSOCIATE FILE-CHARACTER-SET name1 [ name2 ]
          Changes the association for the file character-set name1 to be
          the transfer character-set name2. If name2 is omitted, automatic
          switching is disabled for this file character-set only.

  6.6. UNICODE

   C-Kermit 7.0 adds support for Unicode, the Universal Character Set,
   for:

     * File Transfer (SEND, RECEIVE, GET, etc)
     * Terminal connection (CONNECT)
     * Unguarded file capture (LOG SESSION)
     * Unguarded file transmission (TRANSMIT)
     * Local file character-set conversion (TRANSLATE)

   C-Kermit is not, however, a "Unicode application" in the sense that its
   commands, messages, or user interface are Unicode. Rather, it is
   "Unicode aware" in its ability to handle and convert Unicode text in
   the course of file transfer and terminal connection, and you can also
   use Kermit to convert local files between Unicode and other character
   sets. TLA's:

  BMP - Base Multilingual Plane
  BOM - Byte Order Mark
  CJK - Chinese, Japanese, and Korean
  ISO - International Standards Organization
  TLA - Three-Letter Acronym
  UCS - Universal Character Set
  UTF - UCS Transformation Format

   Unicode and ISO 10646 are the coordinated and compatible corporate and
   international standards for the Universal Character Set (UCS). Unlike
   single-byte and even most multibyte character sets, the UCS can
   represent all characters in every existing writing system. A flat
   plain-text file encoded in some form of UCS can contain any mixture of
   English, Spanish, Italian, German, Hebrew, Arabic, Greek, Russian,
   Armenian, Georgian, Japanese, Chinese, Korean, Vietnamese, Tibetan,
   Hindi, Bengali, Tamil, Thai, Ethiopic, and so on, plus scientific and
   mathematical notation, as well as texts in Runes, Ogham, Glagolitic,
   and other historic scripts.

   The UCS already covers these scripts and many more, but it's an
   evolving standard with efforts underway to accommodate even more
   languages and writing systems. Support is growing for native UCS use on
   many platforms and in many applications. The goal of the framers of the
   UCS is for it to replace ASCII, the ISO Latin Alphabets, ISCII, VISCII,
   the Chinese, Japanese, and Korean (CJK) multibyte sets, etc, as well as
   the many private character sets in use today, in other words to become
   *the* Universal Character Set.

   Until that time, however, conversions between existing sets and the UCS
   will be necessary when moving text between platforms and applications.
   Now Kermit can help.

    6.6.1. Overview of Unicode

   For a more complete picture, please visit:

 http://www.unicode.org/

   and access the various online introductions, FAQs, technical reports,
   and other information. For greater depth, order the latest version of
   the published Unicode Standard. The following overview contains a great
   many oversimplifications and perhaps an opinion or two.

   At present, the UCS is a 16-bit (2-byte) character set, but with
   provisions to grow to a 4-byte set. UCS-2 refers to the two-byte set,
   also called the Base Multilingual Plane (BMP), in which each character
   has 16 bits, and therefore there are 2^16 = 65536 possible characters.
   The first 128 characters are the same as US ASCII (C0 control
   characters and DEL included), the next 32 are the C1 control characters
   of ISO 6429, and the next 96 are the Right Half of ISO 8859-1 Latin
   Alphabet 1. The remaining tens of thousands of characters are arranged
   newly for the UCS, usually (but not always) in sections corresponding
   to existing standards, such as ISO Latin/Cyrillic, often plus
   additional characters not appearing in the existing standards due to
   lack of space (or other reasons).

   ISO 10646 allows for additional planes, e.g. for Egyptian hieroglyphics
   or ancient (or other esoteric) CJK characters, but these planes are not
   yet defined and so we will say nothing more about them here, except
   that their use will require the 4-byte form of UCS, called UCS-4, in
   some form (more about "forms" in Section 6.6.2).

   Unicode and ISO 10646 are constantly under revision, mainly to add new
   characters. The Unicode revision is denoted by a version number, such
   as 1.0, 1.1, 2.0, 3.0. The ISO 10646 standard revision is identified by
   Edition (such as ISO 10646-1 1993), plus reference to any amendments.
   The first versions of these standards included encodings for Korean
   Hangul syllables (Jamos); these encodings were changed in version 1.1
   of Unicode and by Amendment 5 to ISO 10646-1. The Unicode Technical
   Committee and the ISO acknowledge that this was a bad thing to do, and
   promise never change encodings or character names again, since this
   poses serious problems for conformance and data interchange.

   A UCS-2 value is customarily written like this:

  U+xxxx

   where "xxxx" represents four hexadecimal digits, 0-9 and A-F. For
   example, U+0041 is "A", U+00C1 is A-acute, U+042F is uppercase Cyrillic
   "Ya", U+FB4F is Hebrew Ligature Alef Lamed, and U+FFFD is the special
   character that means "not a character".

   Most characters from widely-used alphabetic writing systems such as the
   West European ones, Cyrillic, Greek, Hebrew, Vietnamese, etc, are
   available in "precomposed" form; for example Uppercase Latin Letter A
   with Acute Accent is a single character (as it is in Latin-1). However,
   the UCS also permits composition of a base character with one or more
   nonspacing diacritics. This means the same character can be represented
   in more than one way, which can present problems in many application
   areas, including transfer and character-set conversion of text.

   Conversion from ASCII or Latin-1 to UCS-2 text is "trivial": simply
   insert a NUL (0) byte before each ASCII or Latin-1 byte. Converting in
   the reverse direction (provided the UCS-2 file contains only U+0000 to
   U+00FF) is equally simple (if we ignore the issue of composition):
   remove every second (NUL) byte. Conversion of other character sets to
   and from UCS, however, requires tables or algorithms specific to each
   set. Nevertheless, the relatively transparent upwards compatibility
   from ASCII and Latin-1, in which a very large share of the world's
   textual data is encoded, gives the UCS an entree onto existing
   platforms.

   But the 2-byte format and the preponderance of NUL and other control
   bytes in UCS-2 text pose problems for current applications and
   transmission methods. And to make matters worse, different hardware
   platforms store UCS-2 characters in different byte order. Thus a UCS-2
   file transferred by FTP (or accessed via NFS, etc) between two
   computers with different architecture might have its bytes in the wrong
   order (or worse; see Section 6.6.5.1 ).

    6.6.2. UCS Byte Order

   Consider the number 1. In an 8-bit byte, this would be represented by
   the following series of 0- and 1-bits:

  +-----------------+
  | 0 0 0 0 0 0 0 1 |
  +-----------------+

   Therefore in a 16-bit "word" the representation might be:

  +-----------------+-----------------+
  | 0 0 0 0 0 0 0 0 | 0 0 0 0 0 0 0 1 |
  +-----------------+-----------------+

   Now consider the number 256, which is 2 to the 8th power. The binary
   representation is 100000000 (1 followed by 8 zeros). 256 would go into
   a 16-bit word like this:

  +-----------------+-----------------+
  | 0 0 0 0 0 0 0 1 | 0 0 0 0 0 0 0 0 |
  +-----------------+-----------------+

   When a computer works this way, it is said to be Big Endian, meaning it
   puts the most significant (biggest) byte first (on the "left") in a
   16-bit word, and the least significant byte second (on the right).

   However, some other computers have the opposite arrangement, called
   Little Endian, in which 1 is:

  +-----------------+-----------------+
  | 0 0 0 0 0 0 0 1 | 0 0 0 0 0 0 0 0 |
  +-----------------+-----------------+

   and 256 is:

  +-----------------+-----------------+
  | 0 0 0 0 0 0 0 0 | 0 0 0 0 0 0 0 1 |
  +-----------------+-----------------+

   Computers such as Sparc, MIPS, PA-RISC, and PowerPC are Big Endian,
   whereas the PC and the Alpha are Little Endian. Endianness has never
   been an issue with 7- or 8-bit characters, but it is with UCS
   characters. It can be a tricky business to share or transfer a UCS-2
   file between two different kinds of computers.

   To alleviate (but not entirely solve) the problem, UCS-2 files are
   supposed to begin with the Unicode character U+FEFF, Zero-Width
   No-Break Space (ZWNBS). This is a kind of "no-op" (note: any such
   assertion must normally be qualified with many "but ifs" and "excepts"
   which are omitted here in the interest of brevity). If the bytes are
   reversed the ZWNBS becomes U+FFFE, which is not (and never will be) a
   defined UCS character. U+FEFF at the beginning of a UCS file is
   therefore called a Byte Order Mark, or BOM.

   Any application that creates a UCS-2 (or UTF-16, or UCS-4) file should
   include a BOM, and any application that reads one should test for a
   BOM, and if one is found, infer the byte order from it. This is a
   convention, however -- not a standard or a requirement -- and
   applications vary in their ability to handle BOMs and "backwards" UCS-2
   files.

   Note that a BOM is useful only at the beginning of a file. If you
   append one UCS-2 file to another, and both have BOMs, the internal BOM
   is no longer a BOM. And if the byte orders of the two files differ,
   then either the first part or the second will be backwards. (Various
   other undesirable effects might also occur, not discussed here.)

    6.6.2. UCS Transformation Formats

   UCS textual data can be modified in various ways for transmission or
   storage. Any officially sanctioned method of doing this is called a UCS
   Transformation Format, or UTF. One such method, called UTF-16, is
   essentially identical with UCS-2 except that it designates certain code
   values as "escape sequences" (called surrogate pairs) to access
   characters in other planes without having to use full UCS-4. We won't
   discuss UTF-16 further here, since at the moment there are no other
   planes. Several other UTF's (such as UTF-1, UTF-2, and UTF-7) have
   fallen into disuse and are not discussed here. The most important
   transformation format today is UTF-8.

   UTF-8, so called because it "serializes" UCS-2 data into a stream of
   8-bit bytes, is designed to allow the UCS to work with present-day
   communications gear, computers, and software. The most important
   properties of UTF-8 are that byte order is constant (no byte swapping)
   and all (7-bit) ASCII characters represent themselves. Therefore
   conversion between ASCII and UTF-8 is no conversion at all, and
   applications or platforms (such as Plan 9 from Bell Labs) that use
   UTF-8 "for everything" can still run traditional ASCII-only
   applications and be accessed from them. In particular, unlike UCS-2,
   ASCII characters are not padded with NUL bytes. But also unlike UCS-2,
   there is no transparency for Latin-1 or any other non-ASCII character
   set. Every non-ASCII UCS-2 character is represented by a sequence of 2
   or 3 UTF-8 bytes. Thus UTF-8 is more compact than UCS-2 for text
   containing a preponderance of ABC's (or other ASCII characters), about
   the same as UCS-2 for other alphabetic scripts (Cyrillic, Roman, Greek,
   etc), and larger than UCS-2 for Chinese, Japanese, and Korean.

   The UTF-8 uncoding of the UCS has been adopted by the Internet as the
   preferred character set for new applications, and is gradually being
   retrofitted into traditional applications like FTP ([550]RFC 2640).

    6.6.3. Conformance Levels

   Although the Unicode and ISO 10646 standards both describe the same
   character set, these standards differ in many ways, including their
   stated requirements for conformance and their classification of
   conformance levels.

   Kermit has always abided by ISO character-set standards, including ISO
   character-set designation and invocation methods. In adapting Unicode,
   therefore, we had to choose from among the available ISO designations
   which, in turn, correspond with ISO 10646 conformance levels. At
   present, Kermit claims the lowest conformance level, 1, meaning
   (roughly) that it does not handle combining forms and it does not
   handle Korean Hangul Jamos (just as, at present, it does not handle
   Korean in general). Note that ISO 10646 Conformance Levels 1 and 2
   sidestep the issue of the code changes for Korean Hangul by announcing
   non-support for Hangul regardless of encoding.

   ISO 10646 Conformance Level 1 is approximately equivalent to Unicode
   Normalization Form C (described in Unicode Technical Report 15,
   incorporated into Unicode 3.0).

   As noted in Section 6.6.2, Kermit does not claim to support UTF-16
   at the present time, hence the UCS-2 nomenclature. Kermit treats
   surrogates just as if they were any other UCS-2 characters, rather than
   as escapes to other planes, which means that (except when converting
   between UCS-2 and UTF-8) they are translated to "error" characters,
   since (a) no other planes are defined yet (and if they were, no other
   character sets supported by Kermit would encode their characters), and
   (b) no valid surrogate character corresponds to any other UCS-2
   character.

   A minor yet significant aspect of Unicode 3.0 and some recent
   perturbation of ISO 10646-1 (probably Amendment 18, "Symbols and Other
   Characters") is the addition of the Euro Sign at U+20AC. As noted in
  Section 6.0, Kermit's "Euro compliance" includes conversion
   between Latin Alphabet 9 and various PC code pages. Text can also be
   converted between UCS-2 or UTF-8 and any other Euro-compliant character
   set (Latin-9, CP858, CP1250, CP1252) without loss of the Euro Sign.

    6.6.4. Relationship of Unicode with Kermit's Other Character Sets

   Kermit's character sets are divided into two groups: single-byte sets
   (such as Roman, Hebrew, Cyrillic, Greek) and multibyte (various
   Japanese sets). The two groups are distinct since one normally would
   not expect to convert Kanji ideograms to Roman (or other) letters, or
   vice versa.

   Unicode character-set conversion works with both groups, but obviously
   the result depends on the repertoires of the source and destination
   character-sets both including the characters in the file. For example,
   you can translate a Hungarian text file between Latin-2 and Unicode,
   but not between (say) Unicode and Latin/Greek. By the same token you
   can convert Japanese text from Shift-JIS or EUC or JIS-7 to Unicode and
   back, but you can't convert the same file to (say) Latin-1 if it
   contains Japanese characters.

     JIS-7 is equivalent to DEC Kanji and ISO-2022-JP except that the
     latter two do not support halfwidth Katakana. Kermit treats all
     three of these sets the same way, i.e. as JIS-7.

   As noted, Kermit presently does not handle combining diacritics, and so
   will not correctly convert UCS files that use them into a single-byte
   character set. For example, if a UCS file contains Latin Capital Letter
   A (U+0041) followed by Combining Acute Accent (U+0301), the result will
   be a two-character sequence, A followed by another character. This is
   what is meant by Conformance Level 1. (The situation grows worse with
   multiple diacritics, since they can occur in any order.)

   A higher level of conformance is possible, in which "canonical
   equivalences" are handled via algorithms and databases, at some
   (perhaps considerable) cost in performance, since a fair amount of
   additional code must be executed for every character during data
   transfer (database lookup, sorting of combining sequences into
   canonical order, etc). This can be added in future releases if there is
   a need (but in many cases, pre- and postpostprocessing might be a
   better option).

   Within these constraints, Kermit converts between the UCS and its other
   character sets. For example, a mixture of Russian and English (and/or
   Dutch, or Latin) text can bet converted between the UCS and ISO
   Latin/Cyrillic or KOI-8. But since Kermit does not presently support
   Arabic character-set conversion, the new availability of UCS conversion
   does not mean that Kermit can convert from Arabic UCS text to some
   other character set, because Kermit does not support any other
   character set that includes Arabic. Ditto for Thai, Armenian, Georgian,
   Tibetan, Chinese, Korean, etc. However, Kermit CAN convert Arabic (or
   any other script) between UCS-2 and UTF-8.

   Considering Cyrillic more carefully, note that the UCS also contains
   numerous Cyrillic characters not found in any of the Cyrillic sets (ISO
   Latin/Cyrillic, KOI8, CP866, etc) that Kermit supports; characters
   needed for Abkhasian, Yakut, Tatar, Bashkir, Altaic, Old Church
   Slavonic, etc; UCS text containing any of these historic or "extended"
   Cyrillic characters can not be converted to any of Kermit's current
   single-byte Cyrillic sets without loss. The situation is similar for
   Greek, Hebrew, etc, and even worse for Japanese since Unicode contains
   thousands of Kanjis that are lacking from the Japanese character sets
   based on JIS X 0208, such as EUC-JP, JIS-7, and Shift-JIS.

   In general, when converting from UCS to a single-byte set, there is
   always the possibility of data loss, just as there is when converting
   from any larger set to a smaller one. For example, if a UCS file
   contains Devanagari characters, these characters will be lost when
   converting to (say) Latin-1, just as Roman vowels with acute accents
   are lost when converting from Latin-1 (an 8-bit set) to German ISO 646
   (a 7-bit set).

    6.6.5. Kermit's Unicode Features

   C-Kermit can convert between UCS-2 or UTF-8 and any of its other
   character sets, and also between UCS-2 and UTF-8. When converting
   between UCS-2 or UTF-8 and a non-Unicode character set (such as
   Latin-1), the UCS Line Separator (LS, U+2028) and Paragraph Separator
   (PS, U+2029) characters are converted to the appropriate line
   terminator (CR, LF, or CRLF). When converting from a non-Unicode set to
   UCS-2 or UTF-8, however, line terminators are not converted to LS or
   PS. This is in accordance with the recommendations of Unicode Technical
   Report #13.

   When C-Kermit starts, it tests the native byte order of the computer.
   You can see the result in the SHOW FEATURES or SHOW FILE display. It's
   also available in the variable \v(byteorder): 0 means Big Endian, 1
   means Little Endian.

   When UCS-2 is involved in file transfer or translation, the following
   commands tell C-Kermit what to do about byte order:

   SET FILE UCS BYTE-ORDER { BIG-ENDIAN, LITTLE-ENDIAN }
          This is for reading UCS-2 files that don't have a BOM, and also
          for writing UCS-2 files. If this command is not given, the
          machine's native byte order is used when writing UCS-2 files,
          and also when reading UCS-2 files that don't have a BOM.

   SET FILE UCS BOM { ON, OFF }
          This setting is used when creating UCS-2 files. A BOM is added
          at the beginning by default. Use OFF to not add the BOM. This
          command has no affect when writing files.

   COPY /SWAP-BYTES sourcefile destinationfile
          Use this for fixing a UCS-2 file whose bytes are in the wrong
          order.

   Use SHOW FILE to display the FILE UCS settings.

   Please note, again, that C-Kermit's user interface, including its
   script language, is not internationalized in any way. String
   comparisons, case conversion, and so on, work only for US ASCII
   (comparisons for equality work with other sets, but not
   lexically-greater-or-less-than or caseless comparisons; even
   comparisons for equality can fail when composed characters or byte
   order are involved). String functions such as \findex() and
   \fsubstring() that reference byte positions do just that; they won't
   work with UTF-8 text that contains any non-ASCII characters, and they
   will not work with UCS-2 text at all since they use C strings
   internally, which are NUL-terminated. These are just a few examples to
   illustrate that neither Unicode nor any other character-set beyond
   ASCII is supported at the user-interface, command, or scripting level
   in this version of C-Kermit.

    6.6.5.1. File Transfer

   Kermit supports both UCS-2 and UTF-8 as file and transfer character
   sets in text-mode file transfer.

   To select UCS-2 or UTF-8 as a file character-set, use:

  SET FILE CHARACTER-SET { UCS2, UTF8 }

   If you want to send a UCS-2 text file (or save an incoming file in
   UCS-2 format), tell Kermit to:

  SET FILE CHARACTER-SET UCS2

   and if you want to send a UTF-8 text file (or store an incoming file in
   UTF-8 format), tell Kermit to:

  SET FILE CHARACTER-SET UTF8

   When sending UCS-2 files, Kermit determines the byte order from the
   BOM, if there is one (and if there is a BOM, it is stripped, i.e. not
   sent). If there is no BOM, the byte order is the one specified in the
   most recent SET FILE UCS BYTE-ORDER command, if any, otherwise the
   computer's native byte order is assumed. When storing incoming files as
   UCS-2, the byte order is according SET FILE UCS BYTE-ORDER, if given,
   otherwise the native one; a BOM is written according to SET FILE UCS
   BOM.

   A transfer character-set should be chosen that includes all of the
   characters in the source file. So, for example, if you are sending a
   UCS-2 file containing only German-language text, your transfer
   character-set can be Latin-1, Latin-2, Latin-9, UCS-2, or UTF-8. But if
   you are sending a file that contains a combination of Hebrew and Greek,
   your transfer character-set must be UCS-2 or UTF-8 if you don't want to
   lose one script or the other. Furthermore, the transfer character-set
   must be one that is supported by the receiving Kermit program. Since
   UCS support is new, it is possible that the other Kermit program (if it
   supports character sets at all) does not support it, but does support
   single-byte sets such as Latin-1, Latin/Cyrillic, etc.

   To select UCS-2 or UTF-8 as a transfer character-set, use:

  SET TRANSFER CHARACTER-SET { UCS2, UTF8 }

   It is up to the receiving Kermit program to convert the transfer format
   to its own local format, if necessary. If it does not understand the
   UTF-8 or UCS-2 transfer character-set, and your file can not be
   adequately represented by any single-byte transfer character-set (such
   as Latin-1 or Latin/Cyrillic) then, if UTF-8 format is acceptable on
   the receiving computer, use UTF-8 as the transfer character-set, with
   the receiver told to "set unknown-char keep", or with the sender told
   to "set attribute char off". If you want the file to be stored in UCS-2
   format at the receiver, send it it binary mode if the source file is
   also UCS-2, or else use the TRANSLATE command (next section) to convert
   it to UCS-2 first, then send it in binary mode. You should not use
   UCS-2 as a transfer character-set in text-mode transfers to Kermit
   programs that don't support it, because they are likely to corrupt the
   result the same way FTP would (see the final paragraph of this
   section).

   When UCS-2 is the transfer character set, it always goes into Kermit
   packets in Big Endian format, with no BOM. As always, the transfer
   character-set is announced by the sender to the receiver. The
   announcement for UCS-2 is "I162" (ISO Registration 162 = UCS-2 Level 1)
   and by definition it is Big Endian (the standards say that when UCS-2
   is serialized into bytes, the order must be Big Endian). The
   announcement for UTF-8 is "I190" (UTF-8 Level 1).

   When receiving a file whose transfer character-set is UCS-2 or UTF-8,
   you must choose the appropriate file character set for the result.
   There is no way Kermit can do this for you automatically, since UCS
   data can be in any script at all, or any combination.

   In general, UTF-8 or UCS-2 should be chosen as a transfer character-set
   if the source file is also encoded in some form of UCS and it contains
   more than one script. But there are other situations where where UTF-8
   or UCS-2 offer advantages. For example, suppose the source file is on a
   NeXTstation and the destination file is on VMS. Both the NeXT and the
   DEC Multinational character sets include the French OE digraph, but
   Latin-1 does not. Therefore French words containing this character
   might not arrive intact when Latin-1 is the transfer character-set, but
   will with UTF-8 or UCS-2, since the UCS includes the OE digraph (but so
   does Latin-9).

   UCS-2 should be chosen as a transfer character-set only for Japanese
   text files that contain a large preponderance of Kanji, since in this
   case (and only this case) UCS-2 (two bytes per Kanji) is more efficient
   than UTF-8 (three bytes per Kanji). The same will be true for Chinese
   and Korean when they are supported by Kermit. UCS-2 should never be
   used as a transfer character-set with a transfer partner that does not
   support UCS-2 since this can cause file corruption (see last paragraph
   in this section).

   Note that Kermit's repeat-count compression is 100% ineffective for
   UCS-2, and is also ineffective for multibyte characters in UTF-8 and
   EUC-JP; this is because repeat-compression is a transport-level
   mechanism that operates on a per-byte basis; it has no knowledge of the
   distinction between a byte and a character.

   When C-Kermit starts, it sets up associations ( Section 6.5) for
   incoming files whose transfer character sets are UCS-2 or UTF-8
   appropriately for the platform so that the file character-set for the
   incoming file is UCS-2 in Windows and UTF-8 elsewhere. Otherwise,
   C-Kermit does not make any default associations for UCS-2 or UTF-8, but
   of course you may add or change associations to suit your needs and
   preferences by including the appropriate ASSOCIATE commands in your
   Kermit startup file. For example, if you are a PC user and deal only
   with text written in Greek and English, you can:

  ASSOCIATE TRANSFER-CHARACTER-SET UTF8 CP869
  ASSOCIATE TRANSFER-CHARACTER-SET UCS2 CP869
  ASSOCIATE FILE-CHARACTER-SET CP869 UTF8

   Note that when file transfer involves conversion between a single-byte
   character set and UCS-2 or UTF-8, the file-transfer thermometer and
   estimated time left might be inaccurate, since they are based on the
   source file size, not the transfer encoding. This is purely a cosmetic
   issue and does not effect the final result. (And is not, strictly
   speaking, a bug; Kermit protocol presently includes no method for the
   sender to furnish an "estimated transfer size" to the receiver, and in
   any case any such guess could be as far off as the file size, given the
   many other factors that come into play, such as compression and
   prefixing).

   A caution about FTP and UCS-2. As noted previously, if you transfer a
   UCS-2 file with FTP in binary mode between two computers with opposite
   Endianness, the result will have its bytes in the wrong order. However,
   if you use FTP to transfer a UCS-2 file in "ascii" (text) mode to ANY
   computer, even if it is identical to yours, the result will be
   corrupted because FTP's line-terminator conversions do not account for
   UCS-2. The same holds when sending from a UCS-aware Kermit program to
   an older Kermit program in text mode with a transfer character-set of
   UCS-2. So use UCS-2 as a transfer character-set ONLY with a UCS-2-aware
   Kermit partner.

    6.6.5.2. The TRANSLATE Command

   In Kermit versions that have Unicode support included, TRANSLATE now
   always goes through Unicode; that is, the source set is converted to
   UCS-2 and thence to the target set. This is a major improvement, since
   in prior releases, C-Kermit had to pick the "most appropriate" transfer
   character-set as the intermediate set, and this would result in the
   loss of any characters that the source and target sets had in common
   but were lacking from the intermediate set (for example the OE digraph
   when translating from NeXT to DEC MCS through Latin-1). This never
   happens when Unicode is the intermediate set because Unicode is a
   superset of all other character sets supported by Kermit. A more
   dramatic example would be translation between Cyrillic PC code page 866
   and KOI8-R ( Section 6.4); formerly all the line- and box-drawing
   characters would be lost (since ISO 8859-5 does not have any); now the
   ones that these two sets have in common are preserved.

   UCS-2 and UTF-8 are now both supported as source-file and
   destination-file character sets by C-Kermit's TRANSLATE command, for
   example:

  translate oofa.txt ucs2 latin1 oofa-l1.txt

   translates oofa.txt from UCS-2 to Latin-1, storing the result as
   oofa-l1.txt. Similarly:

  translate oofa.txt utf8 latin1 oofa-l1.txt
  translate oofa.txt latin1 ucs2 oofa-ucs2.txt
  translate oofa.txt latin1 utf8 oofa-utf8.txt
  translate oofa.txt ucs2 utf8 oofa-utf8.txt
  translate oofa.txt utf8 ucs2 oofa-ucs2.txt

   Treatment of the UCS-2 BOM is exactly the same as for file transfer.
   Note that if a UCS-2 source file is in the "wrong" byte order and lacks
   a BOM, and you don't tell Kermit about it with SET FILE UCS BYTE-ORDER,
   the result of the translation is total gibberish. Recall that you can
   use COPY /SWAP-BYTES to switch the byte order of an errant UCS-2 file
   (or any other file for that matter, if you can think of a reason to).
   Also note that:

  translate oofa.txt ucs2 ucs2 new.txt

   Produces a result in the native (or SET FILE UCS) byte-order as long as
   oofa.txt has a BOM.

   As a side benefit of the Unicode work, the TRANSLATE command now works
   for the first time also for all Japanese character sets that Kermit
   supports. In other words, if you have a Japanese text file in any of
   the following encodings:

  EUC-JP
  Shift-JIS
  JIS-7
  UCS-2
  UTF-8

   You can use the TRANSLATE command to convert to any other encoding from
   the same list.

    6.6.5.3. Terminal Connection

   The CONNECT command now allows UTF-8 as a local or remote terminal
   character-set:

  SET TERMINAL CHARACTER-SET { ..., UTF8 } { ..., UTF8 }
  SET TERMINAL REMOTE-CHARACTER-SET { ..., UTF8 }
  SET TERMINAL LOCAL-CHARACTER-SET { ..., UTF8 }

   (Recall that Kermit's terminal character-set has two "ends" -- the set
   used on the host to which Kermit is connected, and the set used on the
   local keyboard and screen.)

   UCS-2 is not supported as a terminal character-set (either end) since
   (a) it is not used that way anywhere to our knowledge, and (b) the
   problems of Endianness and the high likelihood of loss of
   synchronization make it impractical. (Telecommunications is
   byte-oriented; if one byte, or any odd number of bytes, is lost because
   of buffer overruns, circuit resets, etc (or likewise if a burst of
   noise appears that takes the guise of an odd number of bytes), the byte
   order of the subsequent data stream will be backwards; unlike UTF-8 and
   traditional byte-based character sets, UCS-2 is not "self
   synchronizing".)

   UTF-8 does not have byte-order or synchronization problems and is
   growing in popularity as a terminal character set as well as in other
   application areas. It allows a single terminal session to use multiple
   scripts (Roman, Cyrillic, Greek, etc) without ISO 2022 character-set
   switching (which terminal emulators like Kermit 95 can handle but few
   host applications understand or use), and meshes nicely with the
   Unicode screen fonts that are beginning to appear.

   UTF-8 was first used in Plan 9 and soon will be available in Linux. It
   will probably spread from there (Unicode in some form is, of course,
   also used in Windows NT, but only internally -- not for access from
   outside).

   To use UTF-8 or any other character set that uses 8-bit bytes in your
   terminal session, be sure to tell C-Kermit to:

  SET TERMINAL BYTESIZE 8
  SET COMMAND BYTESIZE 8
  SET PARITY NONE

   (or use the shortcut command, EIGHTBIT, which does all three at once).

   In a setup where your local Kermit program uses a single-byte character
   set such as PC Code Page 850 and the remote host uses UTF-8:

  SET TERM CHAR UTF8 CP850

   or:

  SET TERM REMOTE CHAR UTF8
  SET TERM LOCAL CHAR CP850

   all works as expected. UTF-8 text on the remote displays correctly on
   your screen, and when you type CP850 characters, they are translated to
   UTF-8 sequences for transmission, and the echo from the host is
   translated from UTF-8 back to CP850. Telnet negotiations and
   autodownload take place before any character-set translation and work
   as before. The session log (if text mode was selected for it) contains
   only the local terminal character-set. And so on.

   Kermit merely supplies translations from UTF-8 to your local terminal
   character-set (this includes treating UTF-8 Line Separator and
   Paragraph separator as CRLF). However, Kermit does does not, at
   present, perform "canonicalization" of composed sequences, nor does it
   automatically execute bidirectionality algorithms for display of
   mixed-direction text (e.g. Hebrew and English). Such presentation
   issues, like all others in the terminal-host regime, are left to the
   host.

   By the way, C-Kermit also allows UTF-8 to be the local end of the
   terminal character-set, but so far this code is not tested, since we
   don't have a UTF-8 console or terminal to work with. However, it can be
   stated without doubt that C-Kermit's key mapping will not work for
   UTF-8 values, since (a) the key map is indexed by 8-bit byte values and
   (b) C-Kermit reads keystrokes a byte at a time (these comments do not
   apply to K95, which has direct access to the keyboard and can read
   "wide" keycodes and uses them to index a "wide" keymap).

   Restrictions: As noted, the CONNECT command does not support UCS-2 as a
   REMOTE TERMINAL character-set. Neither does it support the Japanese
   sets EUC-JP, JIS-7, and Shift-JIS. Support for the Japanese sets (and
   possibly Chinese and Korean too) might be added in a future release.
   Since the TRANSMIT command (next section) uses the same REMOTE TERMINAL
   character-sets as the CONNECT command, it has the same restrictions.

    6.6.5.4. The TRANSMIT Command

   As described in Chapter 15 ofUsing C-Kermit and Section 4.21
   of this document, the TRANSMIT command can be used to upload a file
   without protocol, more or less as if you were typing it on your
   keyboard while connected to the host. When TRANSMITting in text mode,
   the file's character set is converted to the host's unless you have SET
   TERMINAL CHARACTER-SET TRANSPARENT, or you include the new TRANSMIT
   switch, /TRANSPARENT.

   Before C-Kermit 7.0, the file character-set was assumed to be the same
   as the local end of the terminal character-set, and the TRANSMIT
   command used the same translations as the CONNECT command, ignoring the
   file character-set.

   In C-Kermit 7.0, that assumption (a poor one to begin with) can no
   longer be made, since UCS-2 can be a file character-set but not a
   terminal character-set. So now the file's character-set is given by
   your most recent SET FILE CHARACTER-SET command. The host's character
   set is the remote end of your most recent SET TERMINAL CHARACTER-SET
   command:

  SET TERMINAL CHARACTER-SET remote-set [ local-set ]

   or:

  SET TERMINAL REMOTE-CHARACTER-SET remote-set

   The TRANSMIT command converts each source-file character from the FILE
   character-set to the REMOTE TERMINAL character-set, and then transmits
   the translated characters according to your SET TRANSMIT preferences
   (Chapter 15).

   If you have SET TRANSMIT ECHO ON, and the host is echoing the
   transmitted characters, the echos are converted from the remote
   terminal character-set to the local terminal character-set.

  [ A picture would help... ]

   Confused? Let's work through an example. Suppose your local computer is
   a NeXTstation, on which text files are encoded in the NeXT character
   set, and that the remote computer is a Data General AViiON, which uses
   the Data General International character set. Further suppose that you
   are logged in to the NeXT from a VT220 terminal which uses the DEC
   Multinational character set.

   You need to convert the file from NeXT encoding to DG encoding and
   convert the echoes from DG encoding to DEC encoding. So on the NeXT,
   tell C-Kermit to:

  eightbit
  set file character-set next
  set term character-set dg-international dec-mcs
  transmit /text nextdata.txt

   (This assumes you have some sort of collection process already set up
   on the Data General, such as a text editor or the venerable "cat >
   foo". The EIGHTBIT command is equivalent to SET TERMINAL BYTESIZE 8,
   SET COMMAND BYTESIZE 8, SET PARITY NONE.)

   To further complicate matters, suppose your local terminal character
   set is the same as the remote one, so you don't need terminal
   character-set translation, but you need to TRANSMIT a file that is in a
   different character set and you want it translated to the host set. In
   this case, use SET TERM CHARACTER-SET to actually specify the character
   set used on each end, rather than specifying TRANSPARENT:

  eightbit
  set file character-set ucs2
  set term character-set latin1 latin1
  transmit /text ucs2data.txt

   The distinction between:

  SET TERMINAL CHARACTER-SET xxx yyy

   (where xxx and yyy are the same set) and:

  SET TERMINAL CHARACTER-SET TRANSPARENT

   is new to C-Kermit 7.0, but affects only the TRANSMIT command.

   The TRANSMIT command currently does nothing special with UCS-2/UTF-8
   Line and Paragraph Separator characters; more experience is required to
   find out how these behave in a genuine Unicode terminal-host setting.

   Restrictions: As noted, the TRANSMIT command translates from the FILE
   character-set to the REMOTE TERMINAL character-set. This rules out
   translations to any character set that is not supported as a REMOTE
   TERMINAL character-set, such as UCS-2, EUC-JP, JIS-7, and Shift-JIS.

    6.6.5.5. Summary of Kermit Unicode Commands

   Specifying file character-set and byte order:
          SET FILE CHARACTER-SET { ..., UCS2, UTF8 }
          REMOTE SET FILE CHARACTER-SET { ..., UCS2, UTF8 } (See next
          section)
          SET FILE UCS BOM { ON, OFF }
          SET FILE UCS BYTE-ORDER { BIG-ENDIAN, LITTLE-ENDIAN }

   Specifying the transfer character-set:
          SET TRANSFER CHARACTER-SET { ..., UCS-2, UTF-8 }
          REMOTE SET TRANSFER CHARACTER-SET { ..., UCS-2, UTF-8 }

   Specifying the terminal character-set:
          SET TERMINAL CHARACTER-SET { ..., UTF8 } { ..., UTF8 }
          SET TERMINAL REMOTE-CHARACTER-SET { ..., UTF8 }
          SET TERMINAL LOCAL-CHARACTER-SET { ..., UTF8 }

   Displaying settings:
          SHOW FILE
          SHOW TRANSFER
          SHOW TERMINAL
          SHOW CHARACTER-SETS

   Commands that use these settings include:
          SEND, RECEIVE, GET, etc.
          CONNECT
          TRANSMIT
          LOG SESSION

   Converting files:
          TRANSLATE infile { ..., UCS-2, UTF-8 } { ..., UCS-2, UTF-8 }
          outfile
          COPY /SWAP-BYTES infile outfile

  6.7. Client/Server Character-Set Switching

   A simple mechanism has been added to allow the client to change the
   server's FILE CHARACTER-SET:

   REMOTE SET FILE CHARACTER-SET name
          The client asks the server to change its file character-set to
          the one given. The name must match one of the server's file
          character-set names. For convenience, C-Kermit uses its own file
          character-set keyword list for parsing this command so you can
          use ? for help and Tab or Esc for completion. However, since the
          server might have a different repertoire (or even use different
          names for the same sets), C-Kermit accepts any string you supply
          and sends it to the server. The server, if it supports this
          command (C-Kermit 7.0 and K95 1.1.19 do), sets its file
          character-set as requested, and also disables automatic
          character-set switching ( Section 6.5). If the server does
          not support this command or if it does not support the given
          character set, the REMOTE SET FILE CHARACTER-SET command fails.

   Here's an example that sends a Japanese text file encoded in Shift-JIS
   to a server using every combination of Kermit's Japanese-capable file
   and transfer character sets:

  dcl \&x[] = euc ucs2 utf8             ; transfer character-sets
  dcl \&y[] = eu uc ut                  ; 2-letter abbreviations for them
  dcl \&f[] = shift euc jis7 ucs2 utf8  ; file character-sets
  dcl \&g[] = sj eu j7 uc ut            ; 2-letter abbreviations

  set file char shift-jis               ; local file character-set is Shift-JIS
  for \%i 1 \fdim(&x) 1 {               ; for each transfer character-set...
      set xfer char \&x[\%i]            ; set it
      for \%j 1 \fdim(&f) 1 {           ; for each remote file character-set...
          remote set file char \&f[\%j] ; set it
          if fail exit 1 SERVER REJECTED CHARSET
          send /text meibo-sj.html meibo-sj-\&y[\%i]-\&g[\%j].txt ; send the fil
e
          if fail exit 1 TRANSFER FAILED
      }
  }

   The Kermit-370 server does not support REMOTE SET FILE CHARACTER-SET,
   but since it supports REMOTE KERMIT commands, you can get the same
   effect with REMOTE KERMIT SET FILE CHARACTER-SET name.

  7. SCRIPT PROGRAMMING

   (Also see Section 2.8, Scripting Local Programs.)

  7.0. Bug Fixes

   The following script programming bugs were fixed in C-Kermit 7.0:

     * IF EXIST and IF DIRECTORY were fixed to properly strip braces from
       around their arguments, so "if directory {C:\Program Files}", etc,
       would work as expected. However, this means that if the file or
       directory name is actually enclosed in braces, the braces must be
       doubled.
     * The READ command did not fail if the READ file wasn't open; now it
       does.
     * The READ command refused to read the last or only line of a file if
       it did not end with a proper line terminator; now it does.
     * The END command, when given from within a SWITCH statement, did not
       exit from the current macro or command file; instead it just
       terminated the SWITCH.

  7.1. The INPUT Command

   7.1.1. INPUT Timeouts

   The description of the INPUT command on page 422 fails to mention the
   following two points about the timeout (which apply to C-Kermit 6.0 and
   later):

    1. "INPUT -1 text" (or "INPUT \%x text", where \%x is any variable
       whose value is -1 or less) means "wait forever". This form of the
       INPUT command fails only if it is interrupted, since it will never
       time out.
    2. INPUT 0 performs a nonblocking read of material that has already
       arrived but has not yet been read, and succeeds immediately if the
       target string is found, or fails immediately if it is not found.

   The same points apply to MINPUT. REINPUT ignores its timeout parameter.

    7.1.2. New INPUT Controls

   The following new INPUT controls were added in version 7.0:

   SET INPUT AUTODOWNLOAD { ON, OFF }
          Explained in Section 7.7.

   SET INPUT CANCELLATION { ON, OFF }
          This governs whether an INPUT command can be canceled by
          "pressing any key" on the keyboard. Normally it can be, in which
          case the INPUT command fails immediately and \v(instatus) is set
          to 2, indicating interruption. SET INPUT CANCELLATION OFF
          disables keyboard cancellations; thus if the search text is not
          encountered, the INPUT command will run for its entire timeout
          interval. SET INPUT CANCELLATION OFF does not disable
          interruption by Ctrl-C, however; every command needs an
          emergency exit. (If you really want to disable interruption by
          Ctrl-C, use SET COMMAND INTERRUPTION OFF.)

   Also see Section 7.2 for any new variables related to INPUT.

    7.1.3. INPUT with Pattern Matching

   C-Kermit 7.0 allows INPUT, MINPUT, and REINPUT targets to be a pattern
   (explained inSections 1.19 and4.9). This solves a
   long-standing problem illustrated by the following scenario: a certain
   company has a bank of TCP/IP modem servers, with hostnames server1,
   server2, server3, and so on. Each server's prompt is its name, followed
   by a colon (:), for example "Server72:". Without INPUT patterns, it
   would be rather difficult to wait for the prompt. The brute force
   approach:

  minput 20 Server1: Server2: Server3: ... (enumerating each one)

   is subject to failure whenever a new server is added. A more subtle
   approach:

  input 20 Server
  if fail ...
  input 2 :

   is liable to false positives, e.g. "Welcome to the XYZ Corp Modem
   Server. Please read the following message:"...

   With patterns, you can match the prompt with "Server*:" (which doesn't
   solve the "false positives" problem, but certainly is more compact than
   the brute force method), or with more specific patterns such as
   "Server[1-9]:" and "Server[1-9][0-9]:", or equivalently:

  Server{[1-9],[1-9][0-9]}:

   meaning the word "Server" followed by a single digit (1-9) or by two
   digits representing a number from 1 to 99, followed by a colon.

   INPUT pattern matching has been added in a way that does not interfere
   with existing scripts. No new commands or switches are used. The simple
   rule is: if an INPUT search target is the argument of the (new)
   \fpattern() function, it is a pattern. Otherwise it is taken literally,
   as before. For example:

  input 5 a*b

   searches for an 'a' followed by an asterisk ('*'), followed by a 'b'.
   But:

  input 5 \fpattern(a*b)

   searches for an 'a' followed by anything at all up to and including the
   first 'b'. This means that any search target to INPUT, MINPUT, or
   REINPUT can be a pattern or a literal string, and in particular that
   MINPUT can accommodate any mixture of patterns and literal strings.

   In selecting patterns, note that:

     * A leading '*' is always implied so there is no need to include one.
     * A trailing '*' is meaningless and ignored.
     * A '*' by itself matches the first character that arrives.

   A syntax note: If your pattern is a selection list, meaning a list of
   alternatives separated by commas and enclosed in braces, then the outer
   braces will be stripped by various levels of parsers, so you must
   include three of each:

  input 10 \fpattern({{{abc,mno,xyz}}})

   Note that this is equivalent to:

  minput 10 abc mno xyz

   except for the setting of the \v(minput) variable.

   And a caution: INPUT pattern matching has a limitation that you
   probably never noticed with literal-string matching, namely that there
   is a limit on the size of the match. For example, if the pattern is
   "a*b", the match will succeed if the 'a' and 'b' are not separated by
   more than (say) 8K bytes, but will fail if they are farther apart than
   that. In such cases, it better to use two INPUTs (e.g. "input 10 a" and
   then "input 100 b").

    7.1.4. The INPUT Match Result

   The result of any INPUT, MINPUT, or REINPUT command, no matter whether
   the search targets are patterns or literal strings, is available in the
   new \v(inmatch) variable. For example:

  minput 10 cat \fpattern([dh]og)
  if success echo MINPUT matched "\v(inmatch)"

   This is especially useful when a pattern was matched, since it makes
   the string that matched the pattern available to Kermit; there would be
   no way to get it otherwise.

   After an INPUT command, you can view all the INPUT-related variables by
   typing "show variables in" (abbreviate as "sho var in"), which shows
   the values of all built-in variables whose names start with "in".

  7.2. New or Improved Built-In Variables

   \v(blockcheck)
          Current BLOCK-CHECK setting, 1, 2, 3, or 4. 4 is the code for
          BLANK-FREE-2.

   \v(byteorder)
          The machine's byte order: 0 = Big Endian, 1 = Little Endian.

   \v(cmdbufsize)
          The length of the command buffer, which is the maximum size for
          a macro, a command, a variable, or anything else in C-Kermit's
          script language.

   \v(ctty)
          The device name of C-Kermit's controlling (login) terminal.

   \v(filename)
          Described inSections 4.1 and4.2.

   \v(filenumber)
          Described inSections 4.1 and4.2.

   \v(filespec)
          As of C-Kermit 7.0, contains fully qualified filenames rather
          than (usually) relative ones.

   \v(return)
          Now holds the END n value of the macro that most recently
          returned, in case END was used rather than RETURN.

   \v(editor)
          Pathname of preferred text editor

   \v(editopts)
          Command-line options for editor

   \v(editfile)
          File most recently edited

   \v(browser)
          Pathname of preferred Web browser

   \v(browsopts)
          Command-line options for Web browser

   \v(browsurl)
          URL most recently given to Web browser

   \v(dialtype)
          Type of call most recently placed (see Section 2.1.11).

   \v(kbchar)
          The character, if any, that was typed at the keyboard to to
          interrupt the most recent PAUSE, SLEEP, WAIT, MSLEEP, or INPUT
          command; empty if the most recent such command was not
          interrupted from the keyboard.

   \v(lockdir)
          UNIX only - The name of the UUCP lockfile directory, if known,
          otherwise "(unknown)".

   \v(lockpid)
          UNIX only - PID of process that owns the communication port that
          you tried to open with a SET LINE command that failed because
          the port was in use, otherwise empty. This variable is set with
          every SET LINE command.

   \v(cx_time)
          If no connection (SET HOST, SET LINE, DIAL, TELNET, etc) is
          active, this is 0. If a connection is active, this is the number
          of seconds since the connection was made.

   \v(hwparity)
          If hardware parity is in effect, this variable gives its value,
          such as "even" or "odd" (in which case, the \v(parity) variable
          will be "none"). Otherwise this variable is empty.

   \v(serial)
          Current serial port settings in 8N1 format ( Section 2.10).

   \v(errno)
          In UNIX, the current value of the C runtime errno variable,
          which is quite volatile (meaning that often an "interesting"
          error code can be overwritten by some other library call or
          system service that sets errno before you have a chance to look
          at it). In VMS, the error code returned by the system or library
          call that most recently failed (success codes are not saved).
          Not available in other operating systems.

   \v(errstring)
          The UNIX or VMS system error message that corresponds to
          \v(errno). Not available in all OS's. Also see
         \ferrstring().

   \v(setlinemsg)
          The error message, if any, from the most recent SET LINE, SET
          PORT, SET HOST, TELNET, or other connection-making command. This
          is not necessarily the same as \v(errstring) since these
          commands might fail without generating a system error code, for
          example (in UNIX) because a lockfile existed indicating the
          device was assigned by another user.

   \v(exitstatus)
          The exit status C-Kermit would return if it exited now.

   \v(pexitstat)
          The exit status of the inferior process most recently invoked by
          C-Kermit (by RUN, !, REDIRECT, SEND /COMMAND, etc). In VMS, this
          code can be given to \ferrstring() to get the corresponding
          error message (in UNIX, program/command return codes are not the
          same as system error codes). Not available in operating systems
          other than UNIX and VMS. See Section 4.2.5 for details.

   \v(inmatch)
          The incoming string of characters, if any, that matched the most
          recent INPUT, REINPUT, or MINPUT command.

   \v(intime)
          The number of milliseconds (thousandths of seconds) it took for
          the most recent INPUT command to find its match, or -1 if no
          INPUT command has been given yet. If the INPUT command timed
          out, the value is approximately equal to 1000 times the INPUT
          timeout. If INPUT failed for some other reason, the value is
          undefined (\v(instatus) gives INPUT completion status). If your
          version of C-Kermit is built without high-precision
          floating-point timers, this number will always be a multiple of
          1000.

   \v(inwait)
          The number of seconds specified as the timeout in the most
          recent INPUT command.

   \v(dialsuffix)
          Dialing suffix for use during PDIAL sequence; seeSection
          2.1.10.

   \v(pid)
          UNIX, VMS, and K95 only. C-Kermit's primary process ID, numeric,
          decimal. If you want to show it in hex, use \fn2hex(\v(pid)) If
          you want to show it in octal, use \fn2octal(\v(pid)).

   \v(printer)
          Current printer name or SET PRINTER value.

   \v(p_ctl)
          Control prefix char \v(p_8bit) 8-bit prefix char (if parity not
          none)

   \v(p_rpt)
          Repeat prefix char (if repeat compression enabled)

   \v(herald)
          Kermit's version herald

   \v(test)
          Kermit's test version, if any, or 0 if this is not a test
          version. Typical values for test versions are "Alpha.03" or
          "Beta.14".

   \v(sendlist)
          The number of entries in the SEND-LIST, 0 if none. Note: entries
          do not necessarily correspond to files, since an entry might
          contain wildcards. Also note that the value does not go back to
          0 after the files in the list are sent. To reset this variable,
          use CLEAR SEND-LIST. The purpose of this variable is to
          determine if a SEND command, when given without any filenames,
          will be legal. Example:

  xif \v(sendlist) { send } else { send oofa.txt }

   \v(trigger)
          If the most recent CONNECT session was terminated automatically
          by a trigger, this variable contains the trigger value.

   \v(ty_ln)
          TYPE line number (during TYPE)

   \v(ty_lc)
          TYPE line count (after TYPE)

   \v(ty_mc)
          TYPE match count (after TYPE)

   \v(xferstat)
          Status of most recent file transfer:

-1: No transfer yet
 0: Succeeded
 1: Failed

   \v(xfermsg)
          If the most recent file transfer failed, this is the reason. If
          it succeeded, \v(xfermsg) is an empty string.

   \v(tftime)
          Total elapsed time of most recent file transfer operation, in
          seconds.

   \v(textdir)
          Directory that holds (or is supposed to hold) Kermit text files
          such as installation instructions, release notes, update notes,
          read-me files, "beware" files, etc.

   \v(name)
          The name with which the Kermit program was invoked, e.g.
          "kermit", "wermit", "k95", "k2", etc (see Section 9.1).

   \v(osname)
          Name of operating system on computer where C-Kermit is running,
          obtained at runtime (from uname or equivalent).

   \v(osversion)
          Version of operating system on computer where C-Kermit is
          running, obtained at runtime (from uname or equivalent).

   \v(osrelease)
          Release of operating system on computer where C-Kermit is
          running, obtained at runtime (from uname or equivalent).

   \v(model)
          The specific hardware model of the computer where C-Kermit is
          running, if known.

   \v(math_pi)
          The value of Pi (see Section 7.23)

   \v(math_e)
          The value of e (see Section 7.23)

   \v(math_precision)
          How many significant digits in a floating-point number.

   \v(f_count)
          Result of the most recent FILE COUNT (FCOUNT) command.

   \v(f_error)
          Numeric error code of most recent FILE command.

   \v(f_max)
          Maximum number of files open simultaneously.

   The math constants are given in the precision of underlying computer's
   floating-point arithmetic.

   Note the distinction between \v(osname), \v(osversion), and
   \v(platform); the latter refers to the platform for which and/or upon
   which C-Kermit was built, as opposed to the one on which it is actually
   running. Also note that each operating system can, and probably will,
   interpret and fill in the os* variables differently, or not at all.

   The SHOW VARIABLES command now accepts a variable name, prefix, or
   pattern:

  show variables         Shows all variables.
  show variables t       Shows all variables that start with "t".
  show variables *ver*   Shows all variables whose names contain "ver".
  show variables *ver    Ditto (an implied "*" is appended).

  7.3. New or Improved Built-In Functions

   The following new file-i/o functions are explained inSection
   1.22.

  \f_status(channel)           Status of file open on channel
  \f_pos(channel)              Read/write (byte) pointer of given file
  \f_line(channel)             Current line of file
  \f_handle(channel)           Handle of file
  \f_eof(channel)              Whether given file is at EOF
  \f_getchar(channel)          Read a char from given file
  \f_getline(channel)          Read a line from given file
  \f_getblock(channel,n)       Read a block from given file
  \f_putchar(channel,c)        Write a char to given file
  \f_putline(channel,string)   Write a line to given file
  \f_putblock(channel,string)  Write a block to given file

   The following new date-time-related functions are explained in
  Section 1.6:

  \fday()                Returns day of week of given date
  \fnday()               Returns numeric day of week of given date
  \ftime()               Returns time portion of given date-time
  \fntime()              Converts time to seconds since midnight
  \fn2time()             Converts seconds since midnight to hh:mm:ss
  \fcvtdate(date-time)   Converts free-format date to yyyymmdd hh:mm:ss
  \fdayofyear(date-time) Converts date to yyyyddd (day-of-year) format
  \fdoy(date-time)       Synonym for \fdayofyear()
  \fdoy2date(dayofyear)  Converts yyyyddd to yyyymmdd
  \fmjd(date-time)       Converts free-format date to Modified Julian Date
  \fmjd2date(mjd)        Converts modified Julian date to yyyymmdd

   The new floating-point arithmetic functions are explained in
  Section 7.23. f1 and f2 are floating-point (real) numbers; d is
   the number of decimal places to show:

  \ffpabsolute(f1,d)     Absolute value of f1
  \ffpadd(f1,f2,d)       f1 + f1
  \ffpcosine(f1,d)       Cosine of f1
  \ffpdivide(f1,f2,d)    f1 divided by f2
  \ffpexp(f1,d)          e to the f1 power
  \ffpint(f1)            Integer part of f1
  \ffplog10(f1,d)        Log base 10 of f1
  \ffplogn(f1,d)         Natural log of f1
  \ffpmaximum(f1,f2,d)   Maximum of f1 and f2
  \ffpminimum(f1,f2,d)   Minimum of f1 and f2
  \ffpmodulus(f1,f2,d)   Modulus of f1 and f2
  \ffpmultiply(f1,f2,d)  Product of f1 and f2
  \ffpraise(f1,f2,d)     Raise f1 to power f2
  \ffpround(f1,d)        Round f1 to d places
  \ffpsine(f1,d)         Sine of f1
  \ffpsqrt(f1,d)         Square root of f1
  \ffpsubtract(f1,f2,d)  f2 - f1
  \ffptangent(f1,d)      Tangent of f1

   Integer number functions:

   \fabsolute(n)
          Absolute value of integer n.

   \frandom(n)
          Returns a random integer between 0 and n-1.

   \fradix(s,n1,n2)
          If the string s is an integer in radix n1, the result is the
          same number expressed in radix n2, where n1 and n2 may be any
          number from 2 through 36, expressed as decimal numbers, or
          variables (etc) that evaluate to decimal numbers. For the source
          and result, the digits of any radix, r, are the first r
          characters in the sequence 0-9,a-z (case doesn't matter for the
          letters). The string s may have a sign, + or -; if it starts
          with a minus (-) sign, the result also has a minus sign.

   The \fradix() function does not work with floating-point numbers. It
   does not reveal the internal storage format of a number; for example,
   \fradix(-1,10,16) is -1, not something like FFFFFFFFFF. If all three
   arguments are not given, or if n1 or n2 are not numbers between 2 and
   36 inclusive, or s is not a number in radix n1, an error occurs and the
   empty string is returned. \fradix() also does not offer
   extended-precision arithmetic; number values are limited to those
   expressed as a long integer in the architecture of the underlying
   computer, usually 32 or 64 bits. If you give it an argument whose
   absolute value is larger than can be held in an unsigned long, the
   result is -1.

   The next four are shorthand functions for decimal/hexadecimal and
   decimal/octal number conversion:

   \fn2hex(n)
          Returns the hexadecimal (base 16) representation of the integer
          n. This is different from \fhexify(s), which treats its argument
          as a string rather than a number. The result is always
          left-padded with 0's to make its length even. Examples:

  \n2hex(0)   = "00"                    \fhexify(0)   = "30"
  \n2hex(255) = "ff"                    \fhexify(255) = "323535"
  \n2hex(256) = "0100"                  \fhexify(256) = "323536"

   \fhex2n(x)
          Converts hexadecimal number x to decimal equivalent decimal
          number. This is the inverse of \fn2hex(). Equivalent to
          \fradix(s,16,10).

   \fn2octal(n)
          Returns the octal (base 8) representation of the number n.
          Examples:

  \n2octal(0) = "0"
  \n2oct(255) = "377"
  \n2oct(256) = "400"
  Equivalent to \fradix(n,10,8).

   \foct2n(n)
          Returns the decimal representation of the given octal number, n.
          The inverse of \fn2octal(). Equivalent to \fradix(n,8,10).

   String functions:

   \s(name[n:m])
          Equivalent to \fsubstring(\m(name),n,m) ( Section 7.24).

   \:(name[n:m])
          Equivalent to \fsubstring(name,n,m) (where "name" is any
          \-quantity) ( Section 7.24).

   \fleft(s,n)
          The leftmost ncharacters of string s; equivalent to
          \fsubstring(s,1,n).

   \fstripx(string,char)
          Returns the part of the string up to the rightmost occurrence,
          if any, of the given character. The default character is period
          (.) Examples:

  \fstripx(foo/bar,/)                 = "foo"
  \fstripx(foo/bar/baz,/)             = "foo/bar"
  \fstripx(autoexec.bat,.)            = "autoexec"
  \fstripx(autoexec.bat)              = "autoexec"
  \fstripx(fstripx(foo/bar/baz,/),/)  = "foo"

   \flop(string,character)
          Returns the portion of the string starting after the first
          occurrence of the given character. The default character is
          period (.) Examples:

  \flop(autoexec.bat)                 = "bat"
  \flop(baz.foo/bar)                  = "foo/bar"
  \flop(baz.foo/bar,/)                = "bar

   \fstripn(string,n)
          Returns the string with ncharacters removed from the end.
          Example:

  \fstripn(12345678,3)                = "12345"

          (For more discussion of \fstripx(), \fstripn(), and \flop() see
         Section 4.2.3).

   \fb64encode(s)
          Returns the Base-64 encoding of the string s.

   \fb64decode(s)
          Returns the decoding of the Base-64 string s. Fails if s is not
          a Base-64 string, or if its length is not a multiple of 4. Note
          that if any of the result bytes are null (0), the result string
          stops there. There is no way to represent strings that contain
          null bytes in C-Kermit (the same is true for \funhexify()).

   \fword(s1,n,s2,s3)
          Extracts word number nfrom string s1. By default, a "word" is
          any sequence of ASCII letters or digits; nis 1-based. If n is
          omitted, "1" is used. Examples:

  \fword(one two three)    = "one"
  \fword(one two three,1)  = "one"
  \fword(one two three,2)  = "two"
  \fword(one two three,3)  = "three"

          and:

    \fword(\v(dialresult),2) = "31200"

          is "31200" if \v(dialresult) is (e.g.) "CONNECT
          31200/ARQ/V32/LAPM/V42BIS".

          If you include s2, this replaces the default break set. For
          example, suppose you have a string \%a whose value is:

  $150.00 $300.00 $39.95

          and you want each dollar amount to be a word; use:

  \fword(\%a,\%n,{ })

          This returns dollar amount number \%n, e.g. "$300.00" for \%n =
          2. "{ }" denotes a space (you must enclose it in braces,
          otherwise it is squeezed out). Note that ASCII control
          characters are always included in the break set; you don't have
          to specify them (and you can't not specify them).

          The optional s3 argument lists characters (even control
          characters) that normally would be considered separators that
          you want included in words. So the dollars-and-cents example
          could also be handled this way:

  \fword(\%a,\%n,,$.)

          in other words, use the default separator list, but remove "$"
          and "." from it so they will be considered part of a word.

          Note that since 8-bit characters are not ASCII, they act as
          break characters unless you put them in the include list.
          Suppose, for example, you have a file in which each line is a
          Tab-separated list of words, numbers, or phrases that might
          contain punctuation, special characters like $ and @, 8-bit bit
          characters, etc (like something that might have been exported
          from a spreadsheet or database), and you want to split only on
          Tab; here is a way (\m(line) is a line read from the file):

undef keep
for \%i 1 255 1 {
    if == \%i 9 continue
    .keep := \m(keep)\fchar(\%i)
}
while true {
    fread /line \%c line
    if fail break
    .\%n := \fsplit(\m(line),&a,\9,\m(keep))
    ...
}

          This problem is addressed inC-Kermit 9.0.

   \fsplit(s1,&a,s2,s3)
          This is like \fword(), except instead of extracting and
          returning a particular word from string s1, it counts the words
          and optionally assigns them to the array whose identifying
          letter, a-z, is given after the "&" in the second argument, with
          the first word going into element 1, the second into element 2,
          and so on. The rules regarding break and include lists (s2 and
          s3) are exactly the same as for \fword(). \fsplit() returns the
          number of words that were assigned, which is either the number
          of words in the string, or the dimension of the array, whichever
          is less. If the array is not declared, \fsplit() creates it and
          returns a number which is both the number of words in s1 and the
          dimension of the new array. Examples:

  declare \&w[20]        ; (Optional.)
  ...
  read \%s               ; \%s is "This is a sentence with seven words."
  ...
  echo "\fsplit(\%s)"    ; This would print "7".
  echo "\fsplit(\%s,&w)" ; Ditto, and also assigns them to array \&w[].

  echo "\&w[7]"          ; This would print "words".

          If the line contained fields that were delimited by colon (:),
          you would use \fsplit(\%s,&w,:). If the fields were delimited by
          comma, then you would use \fsplit(\%s,&w,{,}); in this case the
          literal comma must be enclosed in braces to distinguish it from
          the comma that separates function arguments. To get a word count
          without loading an array, but still specify break and/or include
          lists, leave the array argument empty:

   echo "\fsplit(\%s,,:)" ; Use colon as the separator.

          WARNINGS:

         1. If you use the same array repeatedly, \fsplit() leaves any
            trailing members undisturbed. For example:
  dcl \&w[10]
  \fsplit(1 2 3 4 5,&w) ; Sets \&w[1] thru \&w[5].
  \fsplit(a b c,&w)     ; Sets \&w[1]-[3] leaving [4]-[5] as they were.

         2. If you allow \fsplit to create the array (by not declaring it
            first), it is dimensioned to the number of elements it was
            created with:
  \fsplit(1 2 3,&x)     ; Creates an array \&x[] with 3 elements.
  \fsplit(a b c d e,&x) ; This overflows the array.

          Thus if you want to use \fsplit() repeatedly on the same array,
          either dimension it in advance to the maximum expected size (and
          then some -- more efficient), or else destroy it after each use
          (to allow for unexpectedly large arguments). Example using a
          dynamic array:

  fopen /read \%c some-file
  if fail ...
  set function error on    ; See Section 7.12
  while true {
      dcl \&w[]            ; Destroy \&[w] each time thru the loop
      fread /line \%c \%a
      if fail break
      asg \%x \fsplit(\%a,&w)
      if fail ...
      ; (do what you want with \&w[] here...)
  }
  fclose \%c

   \frindex(s1,s2,n)
          The "n" argument to \frindex() now works consistently (in mirror
          image) with the corresponding \findex() argument. In each case,
          the (n-1)-most characters of s2 are ignored in the search; for
          findex, this means the starting position of the search is n (the
          default n is 1, and 0 is treated like 1). For \frindex() it
          means the default starting point is:

  length(s2) - length(s1) - n (with the same defaults for n).

   \fsearch(pattern,string[,position])
          Exactly like \findex(), except with a pattern (seeSection
          7.9) rather than a literal string.

   \frsearch(pattern,string[,position])
          Exactly like \frindex(), except with a pattern rather than a
          literal string.

          File Functions:

   \ffiles(), \fnextfile()
          It is no longer necessary to copy the file list to an array
          before use, as shown on p.398 ofUsing C-Kermit 2nd
          Edition. \ffiles() and friends now make their own safe copies of
          the file list. Thus constructions like the following are now
          possible:

  for \%i 1 \ffiles(*.txt) 1 { send \fnextfile() }

          The same is true for the new function \frfiles(),
          \fdirectories(), and \frdirectories(), described inSection
          4.11.3.

          But note that each reference to \fnextfile() still gets you the
          next file. So "if newer \fnextfile() foo.txt send \fnextfile()"
          compares one file's age with that of foo.txt, and then sends an
          entirely different file. If you're going to refer to the same
          file more than once, assign it to a variable:

  asg \%f \fnextfile()
  if newer \%f foo.txt send \%f

          (note: assign, not define).

          Also note that \ffiles(), \frfiles(), \fdirectories(), and
          \frdirectories() all now accept on optional 2nd argument: the
          name of an array to load with the resulting file or directory
          list, explained in Section 4.11.3. So you can also load an
          array with the filelist when you need to refer to the same file
          more than once:

  for \%i 1 \ffiles(*,&a) 1 { if newer \&a[\%i] foo.txt send \&a[\%i] }

   \fpermissions(file)
          Returns the platform-specific permissions string for the file,
          such as "-rw-rw-r--" in UNIX or "(RWE,RWE,RE,E)" in VMS.

   \fdirname(f)
          Given a file specification f, this function returns the complete
          pathname of directory the file is in.

   Array Functions:

   \fdimension(&a)
          Returns the dimension declared for the array whose identifying
          letter, a-z, or special character "_" or "@", is given after the
          "&" in the argument. If the array is not declared, 0 is
          returned. Note that when used with the macro argument vector
          array, \&_[] (see Section 7.5), the value of this function
          is one less than \v(argc), and when used with the C-Kermit
          command-line argument vector array, \&@[], it is equal to the
          \v(args) variable. Examples:

  echo \fdimension(&a)       ; Not declared.
  0
  declare \&a[12]            ; Now it's declared.
  echo \fdim(&a)
  12

   \farraylook(pattern,arrayname)
          Looks in the given array for the pattern and returns the index
          of the first element that matches, if any, or -1 if none match.
          The arrayname can include a range specifier to restrict to
          search to a segment of the array, e.g.
          \farraylook(*xyz*,&a[32:63]). For greater detail see
         Section 7.10.7.

   \ftablelook(keyword,arrayname[,delimiter])
          Looks in the given "table", which must be sorted, for the given
          keyword. Returns the index of the table element that uniquely
          matches the given keyword, or -1 if none match, or -2 if more
          than 1 match. For greater detail see Section 7.10.7.

   Other new functions:

   \fip2hex(s)
          Converts a dotted decimal IP address to an 8-digit hexadecimal
          number. \fip2hex(128.59.39.2) = 803b2702.

   \fhex2ip(x)
          Converts an 8-digit hexadecimal IP address to dotted decimal
          form, e.g. \fhex2ip(803b2702) = 128.59.39.2. The inverse of
          \fip2hex().

   \fcommand()
   \frawcommand()
          These run an external command and return its output; see
         Section 4.2.8.4.

   \fdialconvert(s)
          s is a phone number in either literal or portable format (not a
          dialing directory entry name). The function returns the dial
          string that would actually be used when dialing from the current
          location (after processing country code, area code, and other
          SET DIAL values).

   \ferrstring(n)
          Returns the system error message associated with the (numeric)
          error code n. UNIX and VMS only. Use in conjunction with
          \v(errno) or \v(pexitstat). See Section 4.2.5 for a usage
          example. Note: This function doesn't work in Windows because
          there is not a consistent error-code-to-message mapping; error
          code "x" means something completely different depending on
          whether it comes from the C runtime library, Winsock, a
          Windows-32 API, TAPI, etc,

   \fpattern(s)
          Used in INPUT, REINPUT, and MINPUT commands to denote search
          strings that are to be treated as patterns rather than
          literally.

   Also see Section 7.8 on built-in help for functions.

  7.4. New IF Conditions

   IF AVAILABLE feature command
          Executes the command if the given feature is available.
          Presently used only to determine if specific authentication and
          encryption options are available. Type "if available ?" to see
          which features may be tested.

   IF FLOAT f1 command
          Executes command if f1 is a legal floating point number (which
          includes integers). Use this to preverify arguments for the
          \ffp...() floating-point arithmetic functions, e.g. "if float
          \%1 echo \ffpint(\%1)".

   IF == n1 n2 command
          Synonym for "if =" (numeric equality). Note that as of C-Kermit
          7.0, this and all other numeric comparison operators also work
          for floating-point numbers.

   IF != n1 n2 command
          Executes the command if n1 and n2 are both numbers or variables
          containing numbers and the value of n1 is not equal to the value
          of n2. This is equivalent to "if not = n1 n2".

   IF <= n1 n2 command
          Executes the command if n1 and n2 are both numbers or variables
          containing numbers and the value of n1 is less than or equal to
          the value of n2. This is equivalent to "if not > n1 n2".

   IF >= n1 n2 command
          Executes the command if n1 and n2 are both numbers or variables
          containing numbers and the value of n1 is greater than or equal
          to the value of n2. Equivalent to "if not < n1 n2".

   IF COMMAND word command
          Executes the command if word is a built-in C-Kermit command.
          Example:

  if not command copy define { copy run copy \%1 \%2 }".

          This defines a COPY macro that runs an external COPY command if
          COPY is not already a built-in command.

   IF LOCAL command
          Executes the command if Kermit is in local mode, i.e. if it has
          a SET LINE, SET PORT, or SET HOST (TELNET, RLOGIN, etc) device
          or connection open. Does not execute the command if in remote
          mode.

   IF MATCH string pattern command
          Executes the command if the string matches the pattern. For a
          description of the syntax for the pattern, seeSection
          4.9.1. If you want to test if the string contains pattern, use
          IF \fsearch(pattern,string).

   IF OPEN { DEBUG-LOG, SESSION-LOG, TRANSACTION-LOG, ... } command
          Executes the command if the given file is open, fails if it is
          not open. Type IF OPEN ? for a complete list of files that can
          be checked (all the files that can be opened with the OPEN or
          LOG commands).

   IF QUIET command
          Executes the command if SET QUIET is ON, and does not execute it
          if SET QUIET is OFF. Example: IF NOT QUIET ECHO { This is a
          message.}.

   IF READABLE name
          Succeeds if name is the name of an existing file or directory
          that is readable.

   IF WRITEABLE name
          Succeeds if name is the name of an existing file or directory
          that is writable, e.g.:

  if not writeable \v(lockdir) echo Please read installation instructions!

   IF FLAG command
          This tests a user-settable condition, which can mean anything
          you like. SET FLAG ON causes subsequent IF FLAG commands to
          succeed; SET FLAG OFF causes them to fail. One way to use it
          would be for debugging your scripts; precede any debugging
          statements with IF FLAG. Then SET FLAG on to debug your script,
          SET FLAG OFF to run it without debugging. Another common use is
          for causing an inner loop to cause an outer loop to exit.

   IF C-KERMIT command
          C-Kermit, but not Kermit 95 or MS-DOS Kermit, executes the
          command.

   IF K-95 command
          Kermit 95, but not C-Kermit or MS-DOS Kermit, executes the
          command.

   IF MS-KERMIT command
          MS-DOS Kermit, but not C-Kermit or Kermit 95, executes the
          command.

  7.5. Using More than Ten Macro Arguments

   The \v(argc) variable now gives the actual number of arguments, even if
   the number is greater than 9:

  C-Kermit> define xx echo \v(argc)
  C-Kermit> xx a b c d e f g h i j k l m n o p q r s t u v w x y z
  27

   Remember that \v(argc) includes the name of the macro itself, so it is
   always at least 1, and is always 1 greater than the actual number of
   arguments. As in versions 6.0 and earlier, if more than 9 arguments are
   given, only the first nine are assigned to the variables \%1..\%9.

   The \&_[] array, discussed on page 353 ofUsing C-Kermit, 2nd ed,
   now holds all the arguments, up to some implementation-dependent limit
   (64 or greater), rather than only the first 9. To illustrate: the
   following macro tells the number of arguments it was called with and
   then prints them:

  define show_all_args {
      local \%i
      echo \&_[0] - Number of arguments: \feval(\v(argc)-1)
      for \%i 1 \v(argc)-1 1 { echo \flpad(\%i,3). "\&_[\%i]" }
  }

   Within a macro \&_[0], like \%0, contains the name of the macro.

   At top level, the \&_[] array is filled as follows:

     * If the first argument on the C-Kermit command line was a filename,
       or C-Kermit was invoked from a "Kerbang" script (Section
       7.19), element 0 contains the filename, and elements 1 through
       \v(argc)-1 hold the remaining command-line arguments.
     * Otherwise the program name goes in element 0, and elements 1
       through \v(argc)-1 hold any arguments that were included after "--"
       or "="

   The new \%* variable, when used within a macro, is replaced by the text
   that followed the macro name in the macro invocation. If no arguments
   were given, \%* is replaced by the empty string. Examples:

  C-Kermit> define xx echo [\%*]
  C-Kermit> define \%a oofa
  C-Kermit> xx
  []
  C-Kermit> xx \%a
  [oofa]
  C-Kermit> xx a
  [a]
  C-Kermit> xx a b
  [a b]
  C-Kermit> xx a b c
  [a b c]
  C-Kermit> xx a b c d e f g h i j k l m n o p q r s t u v w x y z
  [a b c d e f g h i j k l m n o p q r s t u v w x y z]

   Note that \%* can not be used at top level, since Kermit does not have
   access to the raw command line (only to its elements separately, after
   they have been processed by the shell and the C library).

   C-Kermit 7.0 also adds a SHIFT command:

   SHIFT [ number ]
          Shifts the macro arguments (except argument 0) the given number
          of places to the left and adjusts \v(argc) accordingly. The
          default number is 1.

   To illustrate, suppose macro XXX is invoked as follows:

  xxx arg1 arg2 arg3

   Then inside XXX, \%1 is "arg1", \%2 is "arg2", and \%3 is "arg3". After
   a SHIFT command is given inside XXX, then \%1 is "arg2", \%2 is "arg3",
   and \%3 is empty. \%0 (the name of the macro) remains unchanged.

   If more than 9 arguments were given, then arguments are shifted into
   the \%1..9 variables from the argument vector array.

   At top level, the SHIFT command operates on the \&_[] array and \%1..9
   variables; the \&@[] array is not affected. See Section 7.16 for
   details.

   The \%* variable is not affected by the SHIFT command.

  7.6. Clarification of Function Call Syntax

   Spaces are normally stripped from the front and back of each function
   argument; to prevent this enclose the argument in braces:

  \fsplit(\%a,&a,{ })

   However, function calls that contain spaces can make trouble when the
   function is to be used in a "word" field, since space separates words.
   For example:

  for \%i 1 \fsplit(\%a,&a,{ }) 1 {
    echo \%i. "\&a[\%i]"
  }

   In most cases, the trouble can be averted by enclosing the function
   reference in braces:

  for \%i 1 {\fsplit(\%a,&a,{ })} 1 {
    echo \%i. "\&a[\%i]"
  }

   or by replacing spaces with \32 (the ASCII code for space):

  for \%i 1 \fsplit(\%a,&a,\32) 1 {
    echo \%i. "\&a[\%i]"
  }

   Braces are also used in function calls to indicate grouping. For
   example:

  \fsubstring(abcd,2,2) = "bc"

   But suppose "abcd" needed to contain a comma:

  \fsubstring(ab,cd,2,2)

   This would cause an error, since "cd" appears to be the second
   argument, when really you want the first "2" to be the second argument.
   Braces to the rescue:

  \fsubstring({ab,cd},2,2) = "b,"

   Similarly, leading and trailing spaces are stripped from each argument,
   so:

  \fsubstring( abcd ,2,2) = "bc"

   but braces preserve them:

  \fsubstring({ abcd },2,2) = "ab"

   Given these special uses for braces, there is no way to pass literal
   braces to the function itself. For example:

  \fsubstring(ab{cd,2,2)

   causes an error.

   So if you need a function to include braces, define a variable
   containing the string that has braces. Example:

  define \%a ab{cd
  \fsubstring(\%a,2,2) = "b{"

   If the string is to start with a leading brace and end with a closing
   brace, then double braces must appear around the string (which itself
   is enclosed in braces):

  define \%a {{{foo}}}
  \fsubstring(\%a) = "{foo}"

   This also works for any other kind of string:

  define \%a {{ab{cd}}
  echo \fsubstring(\%a) = "ab{cd"

  7.7. Autodownload during INPUT Command Execution

   As of 6.1 / 1.1.12, C-Kermit can be told to look for incoming Kermit
   (or Zmodem) packets during execution of an INPUT command. By default
   (for consistency with earlier releases), this is not done. You can
   enable this feature with:

  SET INPUT AUTODOWNLOAD ON

   (and disable it again with OFF.)

   One possible use for this feature is as a server mode with a time
   limit:

  INPUT 3600 secret-string-to-end-the-INPUT-command

   In this example, any GET, SEND, or REMOTE commands received within one
   hour (3600 seconds) of when the INPUT command was issued will be
   executed. Here's another example, in which we want to stay open until
   11:30pm, or until interrupted by seven consecutive Ctrl-C (\3)
   characters:

  INPUT 23:30:00 \3\3\3\3\3\3\3

   The INPUT AUTODOWNLOAD setting is displayed by SHOW SCRIPTS or SHOW
   INPUT.

  7.8. Built-in Help for Functions.

   Beginning in C-Kermit 7.0, you may obtain a description of the calling
   conventions and return values of any built-in function, such as
   \fsubstring(), with the new HELP FUNCTION command; give the function's
   name without the leading "\f", e.g. "help func substring". You can use
   ?, completion, and abbreviation in the normal manner.

  7.9. Variable Assignments

    7.9.1. Assignment Operators

   Programmers accustomed to languages such as C or Fortran might find
   Kermit's method of assigning values to variables unnatural or awkward.
   Beginning in C-Kermit 7.0, you can use the following alternative
   notation:

 .name = value    is equivalent to   DEFINE name value
 .name := value   is equivalent to   ASSIGN name value
 .name ::= value  is equivalent to   ASSIGN name \feval(value)

   When the command begins with a period (.), this indicates an
   assignment. The name can be a macro name, a \%{digit,letter} variable,
   or an array element. There can be space(s) between "." and the name.
   Examples:

  .\%a = This is a string  ; Same as "define \%a This is a string"
  echo \%a
  This is a string

  .xxx = \%a               ; Same as "define xxx \%a"
  echo \m(xxx)
  \%a

  .xxx := \%a              ; Same as "assign xxx \%a"
  echo \m(xxx)
  This is a string

  declare \&a[2]           ; Use with arrays...
  define \%i 2
  .\&a[1] = first
  .\&a[\%i] = second

   The following sequence illustrates the differences among three levels
   of evaluation:

  .\%x = 2          ; Define a variable to have a numeric value
  .\%y = (3 + \%x)  ; Define another variable as an arithmetic expression

  .xxx = 4 * \%y    ; "=" simply copies the right-hand side.
  echo \m(xxx)
  4 * \%y

  .xxx := 4 * \%y   ; ":=" evaluates the variables first, then copies.
  echo \m(xxx)
  4 * (3 + 2)

  .xxx ::= 4 * \%y  ; "::=" evaluates the expression, then copies.
  echo \m(xxx)
  20

   You can also use this syntax to clear (undefine) a variable:

  .\%a = oofa       ; Define the variable
  echo "\%a"
  "oofa"
  .\%a              ; Clear the variable
  echo "\%a"
  ""

   Extra credit: Can you guess what happens below when the file "abc" does
   not exist?

  fopen /read \%c abc
  if fail ...

    7.9.2. New Assignment Commands

   Recall the DEFINE and ASSIGN commands, and their hidden counterparts,
   _DEFINE and _ASSIGN. The former take the variable name literally, the
   latter evaluate the variable-name field to form the variable name
   dynamically. Examples:

  DEFINE \%x foo    ; Sets the value of the variable \%x to "foo".
  DEFINE \%a \%x    ; Sets the value of the variable \%a to "\%x".
  _DEFINE x_\%a \%x ; Sets the value of the variable x_foo to "\%x".
  ASSIGN \%a \%x    ; Sets the value of the variable \%a to the "foo".
  _ASSIGN x_\%a \%x ; Sets the value of the variable x_foo to "foo".

   This concept has been carried over to the remaining variable-assignment
   commands: EVALUATE, INCREMENT, and DECREMENT:

   EVALUATE variablename expression
          Evaluates the arithmetic expression and assigns its value to the
          variable whose name is given. Example: "eval \%a 1+1" assigns
          "2" to \%a.

   _EVALUATE metaname expression
          Evaluates the arithmetic expression and assigns its value to the
          variable whose name is computed from the given metaname.
          Example: "eval foo<\%a>::\%1 \%2 * (\%3 + \%4)" assigns the
          value of "\%2 * (\%3 + \%4)" to the variable whose name is
          computed from "foo<\%a>::\%1".

   INCREMENT variablename [ expression ]
          Evaluates the arithmetic expression and adds its value to the
          value of the variable whose name is given. Example: "increment
          \%a".

   _INCREMENT metaname [ expression ]
          Evaluates the arithmetic expression and adds its value to the
          value of the variable whose name is computed from the given
          metaname. Example: "_increment Words::\%1.count[\%2]".

   DECREMENT variablename [ expression ]
          Evaluates the arithmetic expression and subtracts its value from
          the value of the variable whose name is given.

   _DECREMENT metaname [ expression ]
          Evaluates the arithmetic expression and subtracts its value from
          the value of the variable whose name is computed from the given
          metaname.

   WARNING: The syntax of the EVALUATE command has changed since C-Kermit
   6.0 and K95 1.1.17. Previously, it did not include a variable name,
   only an expression. To restore the old behavior, use SET EVALUATE OLD.
   To return to the new behavior after restoring the old behavior, use SET
   EVALUATE NEW.

   NOTE: There are no analogs to the "_" commands for the operators
   described in Section 7.9.1; those operators can not be used to
   assign values to variables whose names must be computed.

  7.10. Arrays

   C-Kermit 7.0 adds lots of new array-related features, and groups them
   together under the NEW ARRAY command:

   ARRAY { CLEAR, COPY, DECLARE, DESTROY, RESIZE, SHOW, SORT }

   In each of the ARRAY commands, wherever an array name is expected,
   "short forms" may be used. For example, all of the following are
   acceptable:

  array show \&a[]  (or SHOW ARRAY...)
  array show &a[]
  array show a[]
  array show &a
  array show a

   In addition, ranges are accepted in the ARRAY COPY, ARRAY CLEAR, ARRAY
   SET, ARRAY SHOW, and ARRAY SORT commands:

  array clear \&a[16]     ; Clears 16 thru end
  array clear &a[16]      ; Ditto
  array clear a[16]       ; Ditto

  array clear \&a[16:32]  ; Clears 16 thru 32
  array clear &a[16:32]   ; Ditto
  array clear a[16:32]    ; Ditto

   When using array names as function arguments, you must omit the "\" and
   you must include the "&". You may optionally include empty brackets.
   Examples:

  \fsplit(\%a,a)          ; Bad
  \fsplit(\%a,\&a)        ; Bad
  \fsplit(\%a,&a[3])      ; Bad

  \fsplit(\%a,&a)         ; Good
  \fsplit(\%a,&a[])       ; Good

    7.10.1. Array Initializers

   Beginning in C-Kermit 7.0, you may initialize an array -- in whole or
   in part -- in its declaration:

   [ ARRAY ] DECLARE array-name[size] [ = ] [ value1 [ value2 [...] ] ]

   For compatibility with versions 5A and 6.0, the ARRAY keyword is
   optional. DECLARE can also be spelled DCL.

   Initializers are (a) optional, (b) start with element 1, (c) must be
   enclosed in braces if they contain spaces, and (d) are evaluated
   according to normal rules by the DECLARE command prior to assignment.
   Thus the assignments made here are the same as those made by the ASSIGN
   command. This allows you to initialize array elements from the values
   of other variables. If you actually want to initialize an array element
   to variable's name, as opposed to its value, use double backslashes (as
   in "\\&a", "\\v(time)", etc).

   The size (dimension) of the array is optional. If the size is omitted,
   as in "\&a[]", then the array sizes itself to the number of
   initializers; if there are no initializers the array is not declared
   or, if it was declared previously, it is destroyed. If a size is given,
   any extra elements in the initialization list are discarded and
   ignored.

   NOTE: Unlike in C, the list of initializers is NOT enclosed in braces.
   Instead, braces are used to group multiple words together. So:

  ARRAY DECLARE \&a[] = { one two three }

   would create an array with two elements (0 and 1), with element 1
   having the value " one two three ".

   Examples:

   ARRAY DECLARE \&a[16]
          Declares the array \&a with 17 elements (0 through 16), in which
          all elements are initially empty. If the array \&a[] existed
          before, the earlier copy is destroyed.

   ARRAY DECLARE &a[16]
   ARRAY DECLARE a[16]
   ARRAY DCL \&a[16]
   ARRAY DCL &a[16]
   ARRAY DCL a[16]
   DECLARE \&a[16]
   DECLARE &a[16]
   DECLARE a[16]
   DCL \&a[16]
   DCL &a[16]
   DCL a[16]
          All of the above are the same as the first example.

   ARRAY DECLARE \&a[16] = alpha beta {gamma delta}
          Declares the array \&a with 17 elements (0 through 16),
          initializing \&a[1] to "alpha", \&a[2] to "beta", and \&a[3] to
          "gamma delta". The remaining elements are empty.

   ARRAY DECLARE \&a[] = alpha beta {gamma delta}
          Same as the previous example, but the array is automatically
          dimensioned to 3.

   ARRAY DECLARE \&a[3] = alpha beta {gamma delta} epsilon zeta
          Too many initializers; only the first three are kept.

   ARRAY DECLARE \&a[0]
   ARRAY DECLARE \&a[]
   ARRAY DECLARE &a[]
   ARRAY DECLARE &a
   ARRAY DECLARE a
   DECLARE \&[0]
   DECLARE a
   DCL a
          All of these are equivalent. Each destroys \&a[] if it exists.
          Declaring an array with a dimension of 0 is the same as ARRAY
          DESTROY arrayname.

   ARRAY DECLARE \&a[] = \%1 \%2 \%3
          Declares the array \&a with 3 elements (0 through 3),
          initializing \&a[1] to the value of \%1, \&a[2] to the value of
          \%2, and \&a[3] to the value of \%3. In this case, any reference
          to one of these array elements is replaced by the value of the
          corresponding \%n variable at the time the declaration was
          executed (immediate evaluation; the array element's value does
          not change if the initializer variable's value changes).

   ARRAY DECLARE \&a[] = \\%1 \\%2 \\%3
          Declares the array \&a with 3 elements (0 through 3),
          initializing \&a[1] to the string "\%1", \&a[2] to "\%2", and
          \&a[3] to "\%3". In this case any reference to one of these
          array elements is replaced by the CURRENT value of the
          corresponding \%n variable (deferred evaluation -- the array
          element's value follows the value of the initializer variable).

   The equal sign (=) preceding the initializer list is optional, but is
   recommended for clarity. If you need to initialize element 1 to a
   literal equal sign, use two of them, separated by a space, as in this
   example:

  ARRAY DECLARE \&a[] = = + - * /

   Remember, element 0 is not initialized by the DECLARE command. To
   initialize element 0, use a regular DEFINE or ASSIGN command:

  ARRAY DECLARE \&a[] one two three four five six seven eight nine
  DEFINE \&a[0] zero

   Finally, remember that every command level has its own local array,
   \&_[], containing all the macro arguments (\%0, \%1, ...) for that
   level. See Section 7.5 for details.

    7.10.2. Turning a String into an Array of Words

   The \fsplit(s1,&a,s2,s3) function assigns the words of string s1 to
   successive elements of the array (beginning with element 1) whose
   identifying letter, a-z, is given after the "&" in the second argument,
   using break and include characters given in s2 and s3. SeeSection
   7.3 for details.

    7.10.3. Arrays of Filenames

   See Section 4.11.3 for news about how \ffiles() and related
   functions can assign a list of filenames to an array. To recapitulate
   briefly here:

  \ffiles(*,&a)

   assigns all files that match the first argument to the array denoted by
   the second argument. If the array has not been declared, it is declared
   automatically, with exactly the number of elements needed to hold the
   file list; if it was previously declared, it is destroyed and reused.
   The filenames are assigned starting at array element 1. Element 0 holds
   the number of files in the list.

   The DIRECTORY command ( Section 4.5.1) can also create filename
   arrays if you give it the /ARRAY: switch; this allows selection
   criteria beyond whether the filename matches the given pattern.

   All functions and commands that create filename arrays store the number
   of filenames, n, as element 0 of the array, and the filenames as
   elements 1 through n.

    7.10.4. Automatic Arrays

   In a command file or macro, you can now have local (automatic) arrays.
   Just give the name followed by empty subscript brackets (no spaces
   inside the brackets please) in a LOCAL command, and then declare the
   array:

  LOCAL \%a \&a[] oofa
  ARRAY DECLARE \&a[32] = value1 value2 value3 ...

   This declares the scalar variable \%a, the array \&a[], and the macro
   name "oofa" to be local, and then declares the new local copy of \&a[]
   with 32 elements, perhaps assigning some initial values. When C-Kermit
   exits from the command file or macro containing these command, the
   previous \&a[] array is restored (and if there was no \&a[] at any
   higher level, this will still be true). The process can be repeated to
   any level. Thus it is now safe to write scripts or macros containing
   arrays without danger of interfering with global arrays of the same
   name.

   Just as scalars are inherited by lower command levels, so are arrays.
   So, for example, if \&a[] is declared at top level, all lower levels
   will see it unless they include a "local \&a[]" statement, in which
   case all levels at and beneath the level where the LOCAL statement was
   executed will see the local copy. This too can be repeated to any
   level.

   On the other hand, if you DECLARE an array at a lower command level
   without also making it LOCAL, this replaces the copy that was declared
   at the lowest command level above this one.

    7.10.5. Sorting Arrays

   Although arrays can be sorted using FOR loops as shown on page 383 of
   Using C-Kermit, 2nd Ed., this involves quite a bit of repetitive
   interpretation by the command parser, and so can be slow for large
   arrays. For this reason, C-Kermit 7.0 adds a built-in SORT command:

   ARRAY SORT [ switches ] array [ array2 ]
          Sorts the given array in place. Sorting is strictly lexical
          (string based). The array name can be given fully, e.g. "\&a[]",
          or the "\" and/or "&" and/or brackets can be omitted, e.g.
          "array sort \&a[]", "sort &a", "sort a". Also, a range can be
          indicated in the brackets as noted in Section 7.10, to
          restrict the sort to a range of elements (equivalent to the
          /RANGE switch, described just below), e.g. "array sort
          &a[20:30]".

   A second array may be specified. If it is, and if it is at least as big
   as the first array, it is sorted according to the first array. For a
   sample application, see Section 7.10.10.

   See Section 1.5 for an explanation of switches. The optional
   switches are:

   /CASE:{ON,OFF}
          /CASE:ON means that alphabetic case is significant in
          comparisons; uppercase letters are sorted before lowercase ones.
          /CASE:OFF means case is ignored, e.g. "A" is the same as "a". If
          this switch is not given, sorting is according the current SET
          CASE setting.

   /KEY:n
          Comparison begins at position n(1-based) in each string. If no
          key is given, the entire strings are compared. Only one key can
          be given. If an array element is shorter than the key value, n,
          that element is considered empty for comparison purposes, and
          therefore lexically less than any element at least ncharacters
          long.

   /NUMERIC
          If this switch is included, it means sorting should be numeric,
          rather than lexical. The sort key is the string starting at the
          key position, skipping any leading blanks or tabs, and then as
          much of the string from that point on that fits the definition
          of "numeric", terminating at the first character that does not
          qualify. A numeric string has an optional sign (+ or -) followed
          by one or more digits, and (if your version of Kermit was built
          with floating-point support; see Section 7.23 ) zero or one
          decimal point (period). If both /CASE and /NUMERIC are given,
          /NUMERIC takes precedence.

   /RANGE:n[:m]
          Sort elements nthrough m of the array. By default, the entire
          array from element 1 to its dimensioned size is sorted, which
          might produce surprising results if the array is not full; see
          example in Section 7.10.7. If ":m" is omitted from the
          range, the dimensioned size is used. Thus, to sort an entire
          array, \&a[], including its 0th element, use "sort /range:0 &a".
          You can also sort any desired section of an array, e.g. "sort
          /range:10:20 &a" or "sort /range:\%i:\%j-1 &b". As noted above,
          you can also specify a range in the array-name brackets. If you
          specify a range in the array-name brackets AND with a /RANGE
          switch, the ones in the brackets take precedence.

   /REVERSE
          Sort in reverse order. If this switch is not given, the array is
          sorted in ascending order.

   Remember that numeric switch arguments can be numbers, arithmetic
   expressions, or variables whose values are numbers or expressions, as
   illustrated in the /RANGE examples above.

   A typical sorting application might be to list students' test scores in
   descending order. Suppose you had the following records:

  olaf      65
  olga      98
  ivan      83
  xena     100

   (and so on) stored in array \&s[] (e.g. by reading them from a file as
   illustrated insection 7.10.7). In these records, the student's
   name is in columns 1-9 and the score in 10-12. So to rearrange the list
   in descending order of score:

  sort /key:10 /reverse &s

   Then to list your top five students:

  for \%i 1 5 1 { echo \&s[\%i] }

   Or more simply (see next section):

  show array a[1:5]

   To illustrate the difference between a lexical and a numeric sort,
   suppose you have the following records (the lines that are numbered,
   starting at column 1) in array \&a[]:

    Column   1         2
    12345678901234567890

   1. Ivan 10.0 2. Olaf 9.95 3. Olga 101.5

   ARRAY SORT /KEY:10 &a[] would order them 3,1,2, but ARRAY SORT /KEY:10
   /NUMERIC &a[] would order them 2,1,3.

    7.10.6. Displaying Arrays

   The SHOW ARRAY command (or ARRAY SHOW) now accepts an optional
   array-name argument:

  SHOW ARRAY \&a[]

   (you can leave off the \, the \&, and/or the []'s if you like; "show
   array a" is equivalent to "show array \&a[]"). When an array is
   specified, its dimension is shown and all defined (non-empty) elements
   are listed.

   Example:

  assign \%n \ffiles(*,&a)  ; Fill an array with filenames ( Section 4.11.3)
  show array \&a[]          ; Show the array we just read
  array show \&a[]          ; Same as previous
  array sort \&a[]          ; Sort the array
  array show \&a[]          ; Show it after sorting
  array show \&a            ; Show it again
  array show &a             ; Show it again
  array show a              ; Show it again

   (The final four commands demonstrate the alternative forms that are
   accepted for the array name.)

   If you SHOW ARRAY without giving an array name, all defined arrays are
   listed by name and dimension, but their contents are not shown.

   You can also show a piece of an array by including a subscript or range
   within the array brackets:

  array show \&a[5]         ; Shows \&a[5]
  array show &a[3:8]        ; Shows \&a[3] through \&a[8]
  array show a[:\%n-1]      ; Shows \&a[0] through \&a[\%n-1]

    7.10.7. Other Array Operations

   ARRAY DESTROY arrayname
          Destroys and undeclares the named array. Subscripts or ranges
          are not accepted in this command.

   ARRAY COPY array1 array2
          Copies the first array to the second array. If the target array
          has not been declared, it is created automatically with the same
          size as the first. If it has been declared, it will be used as
          declared; if the source array is larger, only as much of it as
          will fit is copied to the target array. Syntax for array1 and
          array2 is as in ARRAY SHOW (SHOW ARRAY). Example:

  .\%n := \ffiles(*,&a)  ; Create and load array A with a file list.
  array copy &a &b       ; Copy array A to array B.

          The ARRAY COPY command also lets you copy pieces of arrays by
          including range specifiers, as in these examples:

        ARRAY COPY \&a[4:27] \&b
                This copies \&a[] elements 4-27 to \&b[] elements 1-23,
                creating \&b[] if necessary or, if \&b[] is already
                declared, stopping early if its size is less than 23.

        ARRAY COPY \&a[4:27] \&b[12]
                This copies \&a[] elements 4-27 to \&b[] elements 12-35,
                creating \&b[] if necessary or, if \&b[] is already
                declared, stopping early if its size is less than 35.

        ARRAY COPY \&a[4:27] \&b[12:14]
                This copies \&a[] elements 4-6 to \&b[] elements 12-14,
                creating \&b[] if necessary or, if \&b[] is already
                declared, stopping early if its size is less than 14.

        ARRAY COPY \&a[17:] \&b
                This copies all the elements of \&a[] starting with 17
                until the last to \&b[], creating \&b[] if necessary or,
                if \&b[] is already declared, stopping early if \&b[] is
                not big enough. Suppose, for example, you have a script
                whose arguments are string1, string2, and a list of files,
                whose job is to change all occurrences of string1 to
                string2 in each of the files. But if the list of files is
                omitted, then "*" (all files in the current directory) is
                assumed:

if < \v(argc) 3 exit 1 "Usage: \%0 string1 string2 [ files ]"
if not def \%3 {
    .n := \ffiles(*,&a)
} else {
    array copy &_[3:] &a
    .n := \fdim(&a)
}
for i 1 n 1 {
    ! cat \&a[i] | sed -e "s|\%1|\%2|g" > /tmp/_x
    rename /tmp/_x \&a[i]
}

        ARRAY COPY \&a[17] \&b
                Same as previous example.

   ARRAY CLEAR arrayname
          Sets all the elements of the array to the empty value. You may
          also include a range specifier to clear only a selected portion
          of the array; for example "array clear \&a[37:214]". If the
          range is out of bounds, only the part of the array that is in
          bounds is cleared.

   ARRAY SET arrayname [ value ]
          Sets all the elements of the array to the given value. If no
          value is given, the array is cleared. You may also include a
          range specifier to set only a selected portion of the array; for
          example "array set \&a[1:9] -1". If the range is out of bounds,
          only the part of the array that is in bounds is set.

   ARRAY RESIZE arrayname size
          Resizes the given array. If the size is greater than the array's
          current dimension, new empty elements are added to the end. If
          the size is less than the current dimension, the extra elements
          are discarded. Note: If you have stored the array size in
          element 0, ARRAY RESIZE does not change this value. Alternative
          notation: ARRAY RESIZE arrayname[size]. For a practical example,
          see Section 7.10.11.

   \farraylook(pattern,arrayname)
          This function returns the index of the first element of the
          given array that matches the given pattern (for details about
          pattern syntax, seesection 4.9). The array name can
          include a range specification to restrict the search to a given
          segment of the array. If no elements match the pattern, -1 is
          returned.

   \ftablelook(keyword,arrayname[,delimiter])
          Looks in the given "table", which must be sorted, for the given
          keyword. The keyword need not be spelled out in full.
          Pattern-matching characters should not be included as part of
          the keyword. The function returns the index of the table element
          that uniquely matches the given keyword, or -1 if none match, or
          -2 if more than 1 match.

   A "table" is an array that is sorted in lexical order; each of its
   elements may contain multiple fields, delimited by the given delimiter
   character or, if no delimiter is specified, a colon (:).

   The \farraylook() function does exactly what you tell it. If you give
   it a pattern that does not include wildcard characters (such as *, ?,
   etc), it requires an exact match. For example:

  \farraylook(oofa,&a)

   searches for the first element of \&a[] whose value is "oofa". But:

  \farraylook(oofa*,&a)

   finds the first element whose value starts with "oofa", and;

  \farraylook(*oofa,&a)

   finds the first element whose value ends with "oofa", and;

  \farraylook(*oofa*,&a)

   finds the first element whose value contains "oofa".

   Here's a simple demonstration of looking up patterns in arrays:

  local \&a[] \%x \%n
  declare \&a[] = zero one two three four five six seven eight nine ten
  while true {
      .\%x = 1
      .\%n = 0
      ask \%a { Pattern? }
      if not def \%a exit 0 Done.
      while <= \%x \fdim(&a) {
          .\%x := \farraylook(\%a,&a[\%x])
          if ( < \%x 0 ) break
          echo \flpad(\%x,3). \&a[\%x]
          increment \%x
          increment \%n
      }
      if ( < \%n 1 ) echo Pattern not found - "\%a"
  }

   The array need not be sorted. When a pattern is given, a search is
   performed; if there is a match, the matching element's index and the
   element itself are printed, and the search begins again at the next
   element. Thus each matching element is printed. If none match, the
   "Pattern not found" message is printed. The process repeats for as many
   patterns as the user wants to type, and terminates when the user types
   an empty pattern.

   Now let's build a little command parser, consisting of a keyword table,
   and a loop to look up the user's commands in it with \ftablelook(). In
   this case the array elements have "fields" separated by colon (:) -- a
   keyword and a value. Keyword tables must be sorted if \tablelook() is
   to work right, so after declaring and initializing the table array, we
   sort it.

  local \&k[] \%a \%i \%n

  array declare \&k[] = drive:9 do:8 discuss:7 live:6 spend:5 help:4 quit:0

  array sort &k                             ; Make sure array is sorted
  echo Type "help" for help.                ; Print greeting & instructions

  while true {                              ; Loop to get commands
      undefine \%a
      while not defined \%a {               ; Get a command
          ask \%a { Command? }
      }
      .\%n := \ftablelook(\%a,&k)           ; Look up the command
      switch \%n {                          ; Handle errors
        :-1, echo Not found - "\%a"         ; Doesn't match
             continue
        :-2, echo Ambiguous - "\%a"         ; Matches too many
             continue
      }
      switch \fword(\&k[\%n],2) {           ; Dispatch according to value
         :9, echo Driving..., break
         :8, echo Doing..., break
         :7, echo Discussing..., break
         :6, echo Living..., break
         :5, echo Spending..., break
         :4, echo { Commands (may be abbreviated):}
             for \%i 1 \fdim(&k) 1 {
                echo {  \%i. \fword(\&k[\%i],1) }
             }
             break
         :0, exit 0 Bye!
         :default, stop 1 Internal error
      }
  }

   In this example, keywords are "drive", "do", "discuss", etc, and their
   values are unique numbers (values need not be numbers, and there need
   not be only one value -- there can be 0, 1, 2, or more of them). The
   user types a command, which can be the whole word (like "help") or any
   abbreviation (like "hel", "he", or just "h"). If this does not match
   any keywords, \ftablelook() returns -1; if it matches more than one (as
   would "d"), it returns -2. Otherwise the array index is returned, 1 or
   higher.

   Given the array index \%n, we can get the table values as follows:

  \fword(\&k[\%n],1) is the keyword (first field)
  \fword(\&k[\%n],2) is the value (second field, in this case a number)

   In our example, we use the value (number) as the SWITCH variable. As
   noted, \fablelook() expects the array elements to contain multiple
   fields separated by colon (:) (or other character that you specify,
   e.g. \ftablelook(\%a,&a,^)) and when matching the keyword, ignores the
   first delimiter and everything after it.

    7.10.8. Hints for Using Arrays

   C programmers are accustomed to out-of-bounds array references causing
   core dumps or worse. In C-Kermit:

     * A reference to an an out-of-bounds array element returns the empty
       string.
     * An attempt to set the value of an array element that is out of
       bounds or that has not been declared simply fails.

   C programmers expect an array of size n to have elements 0 through n-1.
   Fortran programmers expect the same array to have elements 1 through n.
   C-Kermit accommodates both styles; when you declare an array of size n,
   it has n+1 elements, 0 through n, and you can use the array in your
   accustomed manner, 0-based or 1-based.

   However, note that C-Kermit has certain biases towards 1-based arrays:

     * Assignment of file lists starts with element 1 (Section
       7.10.3).
     * Assignment by \fsplit() starts with element 1 ( Section 7.3).
     * Array initialization skips the 0th element. To initialize a 0-based
       array, use something like this:
  declare \&a[3] = one two three
  .\&a[0] = zero

     * The ARRAY SORT command skips the 0th element unless you include
       /RANGE:0
     * The SHIFT command ignores element 0 of the \&_[] array.

   The distinction between an array's dimensioned size and the number of
   elements in the array is important when sorting. To illustrate:

  declare \&a[100]                  ; Declare array &a with 100 elements
  fopen /read \%c oofa.txt          ; Open a file
  if fail...
  for \%i 1 \fdim(&a) 1 {           ; Read the file into the array
      fread \%c \&a[\%i]
      if fail break
  }
  fclose \%c
  if > \%i \fdim(&a) end 1 File has too many lines for array.
  .\%n ::= \%i - 1
  echo File has \%n line(s).

   Let's say the file had 95 lines. This leaves elements 96-100 of the
   array empty. Now suppose you sort the array and write out the result:

  sort &a                           ; Sort the whole array
  fopen /write \%o oofa.txt.sorted  ; Open an output file
  if fail ...
  for \%i 1 \%n 1 {                 ; Write out 95 records
      fwrite /line \%o \&a[\%i]
      if fail end 1 Write error
  }
  close write

   You might be surprised at the contents of "oofa.txt.sorted" -- five
   empty elements, 96-100, floated to the top of the array in the sort,
   and since your write loop only had 95 iterations, the final 5 lines of
   the sorted file are lost.

   Therefore, when dealing with partially filled arrays -- especially when
   sorting them -- remember to specify the number of elements. A handy way
   of recording an array's "true" size is to put it in the 0th element.
   That way, it "travels with the array". To illustrate (continuing the
   previous example at the "close read" statement):

  close read
  if > \%i \fdim(&a) end 1 File has too many lines for array.
  .\&a[0] ::= \%i - 1     ; Assign number of lines to \&a[0].
  echo File has \&a[0] line(s).
  sort /range:1:\&a[0] &a
  open write oofa.txt.sorted
  if fail ...
  for \%i 1 \&a[0] 1 {
      writeln file \&a[\%j]
      if fail end 1 Write error
  }
  close write

   Note the SORT switch, /RANGE:1:\&a[0]. This keeps the sort 1-based, and
   uses element 0 of the array as its size indicator.

   Finally, note that even though some commands or functions might put a
   size in array element 0, no built-in functions or commands depend on a
   size actually being there. Thus you are perfectly free to replace the
   size with something else and treat the array as 0-based.

    7.10.9. Do-It-Yourself Arrays

   Kermit's \&x[] arrays are nice because of the accompanying built-in
   functionality -- ARRAY commands, built-in functions that load and
   search arrays, automatic evaluation of arithmetic expressions within
   the subscript brackets, and so on. Yet they also have certain
   limitations:

    1. Except when created by dynamic loading (e.g. by \ffiles()) they
       must be declared and dimensioned in advance.
    2. Indices must be numeric, positive, and in range.
    3. There can be only one dimension. Matrices or other
       higher-dimensioned arrays are not available.

   But none of this is to say you can't invent any kind of data structure
   you like. In Section 7.9.2 you can see some examples. Here's
   another (courtesy of Dat Thuc Nguyen), in which a pair of matrices is
   created and then added: no dimensioning necessary.

  .row = 4
  .col = 9

  ; MACRO TO PRINT A MATRIX
  define PMATRIX {
      echo Matrix \%1:
      for \%r 1 \m(row) 1 {
          for \%c 1 \m(col) 1 {
              xecho \flpad(\m(\%1[\%r][\%c]),4)
          }
          echo
      }
      echo
  }
  ; CREATE MATRICES A AND B
  for \%r 1 \m(row) 1 {
      for \%c 1 \m(col) 1 {
          _eval A[\%r][\%c] \%r + \%c
          _eval B[\%r][\%c] \%r * \%c
      }
  }
  ; CREATE MATRIX C = SUM OF MATRIX A AND MATRIX B
  for \%r 1 \m(row) 1 {
      for \%c 1 \m(col) 1 {
          _eval C[\%r][\%c] \m(A[\%r][\%c]) + \m(B[\%r][\%c])
      }
  }
  pmatrix A  ; Print Matrix A
  pmatrix B  ; Print Matrix B
  pmatrix C  ; Print Matrix C

   In the example, we use matrix-like notation to create macros with names
   like "A[1][1]", "B[3][7]", and so on.

    7.10.10. Associative Arrays

   An associative array is a special kind of Do-It-Yourself array. It
   differs from a regular array in that its indices need not be numbers --
   they can be anything at all -- words, filenames, names of months, any
   character string at all, and that it doesn't have to be (and in fact
   can't be) declared. An associative array element is simply a macro
   whose name ends with an index enclosed in angle brackets, for example:

  file<oofa.txt>

   More formally:

  basename<index>

   An associative array is a collection of all associative array elements
   that have the same basename. Any number of associative arrays, each
   with any number of elements, can exist at the same time.

   An associative array element can be assigned a value, such as "1", just
   like any other macro:

  define file<oofa.txt> 1     ; Give "file<oofa.txt>" the value "1".

   or:

  assign file<oofa.txt> \%a   ; Give it the value of the variable \%a.

   However, since an associative array element is a macro, it may not have
   an empty (null) value, since assigning an empty value to a macro
   undefines the macro.

   You can refer to the value of an associative array element using the
   familiar notation for macro values:

  echo \m(file<oofa.txt>)     ; Echo the value of "file<oofa.txt>".

   Associative arrays are most useful, however, when the value of the
   index is a variable. In that case, you must use the "hidden" forms of
   the DEFINE or ASSIGN commands that evaluate the macro name before
   making the assignment (seeUsing C-Kermit, page 457). Example:

  define \%f oofa.txt
  _define file<\%f> 1
  echo file<\%f> = \m(file<\%f>)

   prints:

  file<oofa.txt> = 1

   and then:

  _increment file<\%f>
  echo file<\%f> = \m(file<\%f>)

   prints:

  file<oofa.txt> = 2

   What are associative arrays good for? The classic example is "word
   counts": finding the number of times each word is used in a text
   without knowing in advance what the words are. Without associative
   arrays, your program would have to build a table of some kind, and
   every time a word was encountered, look it up in the table to find its
   position and counter, or add it to the table if it wasn't found -- a
   time-consuming and laborious process. Associative arrays, however, let
   you use the word itself as the table index and therefore sidestep all
   the table building and lookups.

   Let's work through a practical example. Suppose you have a
   file-transfer log in which each line is composed of a number of
   blank-separated fields, and the 9th field is a filename (which happens
   to be the format of certain FTP server logs, as well as of C-Kermit's
   new FTP-format transaction log, described in Section 4.17.2), for
   example:

  Wed Jul 14 09:35:31 1999 22 xx.mit.edu 13412 /pub/ftp/mm/intro.txt ....

   and you want to find out how many times each file was transferred. The
   following code builds an associative array, file<>, containing the
   counts for each file:

  local name line max \%c \%n          ; Declare local variables
  fopen /read \%c /var/log/ftpd.log    ; Open the log file ( Section 1.22)
  if fail exit 1 Can't open log        ; Check
  while true {                         ; Loop for each record
      fread /line \%c line             ; Read a line
      if fail break                    ; Check for end of file
      .name := \fword(\m(line),9,{ })  ; Get 9th field = filename (Sec 7.3)
      _increment file<\m(name)>        ; Increment its counter (Sec 7.9.2)
  }
  fclose \%c                           ; Close file when done.

   Note that _INCREMENT (and INCREMENT, and [_]DECREMENT) treat an empty
   (i.e. nonexistent) variable as having a value of 0, and therefore
   creates the variable with a value of 1.

   At this point, if you told Kermit to "show macro file<", it would list
   the associative array. But since you don't necessarily know the names
   of the files in the array, or even how many elements are in the array,
   how can you use it in a script program?

   The idea of creating macro names that include character-string indices
   enclosed in angle brackets is perfectly arbitrary and doesn't depend on
   any Kermit features that weren't already there -- we could just as
   easily have used some other notation, such as "file[index]",
   "file:index", or "file.index", and the code above would have worked
   just as well (with the corresponding syntax adjustments). But to be
   able to use an associative array in a program after the array is built,
   we need a method of accessing all its elements without knowing in
   advance what they are. That's where the chosen notation comes in.

   First of all, any macro name that ends with "<xxx>" (where "xxx" is any
   string) is case sensitive, unlike all other macro names, which are case
   independent. To illustrate, "file<oofa.txt>" and "file<OOFA.TXT>" are
   two distinct macros, whereas "OOFA", "Oofa", and "oofa", when used as
   macro names, are all the same.

   Second, the new \faaconvert() function converts an associative array
   (that is, all macros with names of the form "base<index>" that have the
   same "base" part) into a pair of regular arrays and returns the number
   of elements:

  \faaconvert(name,&a[,&b])

   "name" is the name of the associative array, without the angle brackets
   or index ("file" in our example).

   The second argument is the name of a regular array in which to store
   the indices of the associative array (filenames in our example); if an
   array of this name already exists, it is destroyed unless the array is
   LOCAL. The third argument is the name of another regular array in which
   to store the values (the counts in our example), with the same rules
   about array name collisions. If you care only about the indices and not
   the values, you can omit the third argument to \faaconvert(). In any
   case, the associative array is converted, not copied: its elements are
   moved to the specified regular arrays, so after conversion the original
   associative array is gone.

   As with other array-loading functions, \faaconvert() sets element 0 of
   each array to the number of elements in the array.

   To continue our example:

  .max := 0                                   ; Maximum count
  .\%n := \faaconvert(file,&a,&b)             ; Convert
  for \%i 1 \%n 1 {                           ; Loop through values
      echo \flpad(\%i,3). \&a[\%i]: \&b[\%i]  ; Echo this pair
      if ( > \&b[\%i] \m(max) ) {             ; Check for new maximum
          .name := \&a[\%i]
          .max  := \&b[\%i]
      }
  }
  echo Most popular file: \m(name), accesses: \m(max)

   This lists the files and counts and then announces which file has the
   highest count.

   Now suppose you want to sort the array pair created from an associative
   array. In our example, \&a[] contains filenames, and \&b[] contains the
   associated counts. Here we take advantage of the ARRAY SORT command's
   ability to sort a second array according to the first one:

  array sort /reverse /numeric &b &a          ; Descending sort by count

   Now to see the top five files and their counts:

  echo The top 5 files are:
  for \%i 1 5 1 {                             ; Loop through top 5 values
      echo \flpad(\%i,3). \&a[\%i]: \&b[\%i]  ; Echo this pair
  }

    7.10.11. Transferring Array Contents to Other Computers

   The SEND /ARRAY:arrayname command ( Section 4.7.1) allows you to
   send the contents of any array, or any contiguous segment of it, in
   either text or binary mode to another computer, using Kermit protocol.
   When used in conjunction with C-Kermit's other features (the array
   features described in this section; the file i/o package from
  Section 1.22; its decision-making, pattern-matching, and string
   manipulation capabilities, and so on) the possibilities are endless:
   extracts of large files, remote database queries, ..., all without
   recourse to system-dependent mechanisms such UNIX pipes and filters,
   thus ensuring cross-platform portability of scripts that use these
   features.

   When sending an array in text mode, Kermit appends a line terminator to
   each array element, even empty ones, and it also converts the character
   set from your current FILE character-set to your current TRANSFER
   character-set, if any. No conversions are made or line terminations
   added in binary mode. For example, the following array:

  dcl \&a[] = One Two Three Four Five Six

   is sent as six lines, one word per line, in text mode, and as the bare
   unterminated string "OneTwoThreeFourFiveSix" in binary mode.

   You should always include a /TEXT or /BINARY switch in any SEND /ARRAY
   command to force the desired transfer mode, otherwise you're likely to
   be surprised by the effects described in Section 4.3.

   Here are some examples:

   send /text /array:\&a[]
          Sends the entire contents of the array \&a[] in text mode. Since
          an as-name is not included, the receiver is told the filename is
          _array_a_.

   send /text /array:&a[]
   send /text /array:a[]
   send /text /array:&a
   send /text /array:a
          These are all equivalent to the previous example.

   send /text /array:&a /as-name:foo.bar
          As above, but the array is sent under the name foo.bar.

   send /text /array:&a[100:199] /as:foo.bar
          As above, but only the elements from 100 through 199 are sent.

   In text-mode transfers, character sets are translated according to your
   current settings, just as for text files. In binary mode, of course,
   there is no character-set translation or other conversion of any kind.
   But remember that array elements can not contain the NUL (ASCII 0)
   character, since they are implemented as NUL-terminated strings.

   Here's an example that shows how to send all the lines (up to 1000 of
   them) from a file animals.txt that contain the words "cat", "dog", or
   "hog" (see Section 4.9 about pattern matching):

  declare \&a[1000]
  fopen /read \%c animals.txt
  if fail exit 1
  .\%i = 0
  while true {
      fread \%c line
      if fail break
      if match {\m(line)} {*{cat,[dh]og}*} {
          increment \%i
          if ( > \%i \fdim(&a) ) break
          .\&a[\%i] := \m(line)
      }
  }
  fclose \%c
  send /array:a[1:\%i] /text

   Note that we are careful to send only the part of the array that was
   filled, not the entire array, because there are likely to be lots of
   unused elements at the end, and these would be sent as blank lines
   otherwise.

   This example raises an interesting question: what if we want to send
   ALL the matching lines, even if there are more than 1000 of them, but
   we don't know the number in advance? Clearly the problem is limited by
   Kermit's (and the computer's) memory. If there are a thousand trillion
   matching lines, they most likely will not fit in memory, and in this
   case the only solution is to write them first to a temporary file on
   mass storage and then send the temporary file and delete it afterwards.

   However, when the selection is likely to fit in memory, the
   once-familiar technique of initial allocation with extents can be used:

  if match {\m(line)} {*{cat,[dh]og}*} {
      increment \%i
      if ( > \%i \fdim(&a) ) {
          array resize a \fdim(&a)+100
          if fail stop 1 MEMORY FULL
          echo NEW DIMENSION: \fdim(&a)
      }
      .\&a[\%i] := \m(line)
  }

   This grows the array in chunks of 100 as needed.

  7.11. OUTPUT Command Improvements

   LINEOUT [ text ]
          This command is exactly like OUTPUT, except it supplies a
          carriage return at the end of the text. "lineout exit" is
          exactly the same as "output exit\13".

   SET OUTPUT SPECIAL-ESCAPES { ON, OFF }
          This command lets you tell C-Kermit whether to process \N, \L,
          and \B specially in an OUTPUT command, as distinct from other \
          sequences (such as \%a, \13, \v(time), etc). Normally the
          special escapes are handled. Use SET OUTPUT SPECIAL-ESCAPES OFF
          to disable them.

   Disabling special escapes is necessary in situations when you need to
   transmit lines of data and you have no control over what is in the
   lines. For example, a file oofa.txt that contains:

  This is a file
  It has \%a variables in it
  And it has \B in it.
  And it has \L in it.
  And it has \N in it.
  And this is the last line.

   can be sent like this:

  local line
  set output special-escapes off
  fopen /read \%c oofa.txt
  if fail stop 1 Can't open oofa.txt
  while success {
      fread \%c line
      if fail break
      ; Add filtering or processing commands here...
      output \m(line)\13
  }

  7.12. Function and Variable Diagnostics

   In C-Kermit 6.0 and earlier, the only diagnostic returned by a failing
   function call was an empty value, which (a) could not be distinguished
   from an empty value returned by a successful function call; (b) did not
   give any indication of the cause of failure; and (c) did not cause the
   enclosing statement to fail. C-Kermit 7.0 corrects these deficiencies.

   SET FUNCTION DIAGNOSTICS { ON, OFF }
          when ON, allows built-in functions to return diagnostic messages
          when improperly referenced, instead of an empty string. FUNCTION
          DIAGNOSTICS are ON by default. When OFF, improperly referenced
          functions continue to return an empty string. This command also
          affects built-in variables; in this case, an error message is
          returned only if the variable does not exist. When FUNCTION
          DIAGNOSTICS are ON, the error message is also printed.

   For variables, the only message is:

  <ERROR:NO_SUCH_VARIABLE:\v(name)>

   where "name" is the name of the nonexistent variable.

   For functions, the diagnostic message is:

  <ERROR:message:\fname()>

   where "message" is replaced by a message, and "name" is replaced by the
   function name, e.g. <ERROR:ARG_NOT_NUMERIC:\fmod()>. Messages include:

  ARG_BAD_ARRAY       An argument contains a malformed array reference.
  ARG_BAD_DATE        An argument contains a malformed date and/or time.
  ARG_BAD_PHONENUM    An argument contains a malformed telephone number.
  ARG_BAD_VARIABLE    An argument contains a malformed \%x variable.
  ARG_INCOMPLETE      An argument is incomplete (e.g. a broken Base64 string).
  ARG_EVAL_FAILURE    An argument could not be evaluated (internal error).
  ARG_NOT_ARRAY       An argument references an array that is not declared.
  ARG_NOT_NUMERIC     An argument that must be integer contains non-digits.
  ARG_NOT_FLOAT       An argument has bad floating-point number format.
  ARG_NOT_VARIABLE    An argument that must be a variable is not a variable.
  ARG_OUT_OF_RANGE    An argument's numeric value is too big or too small,
                      or an argument contains illegal characters (e.g. a hex
                      or Base-64 string).
  ARG_TOO_LONG        An argument's value is too long.
  ARRAY_FAILURE       Failure to create an array.
  DIVIDE_BY_ZERO      Execution of the function would cause division by zero.
  FLOATING_POINT_OP   Execution error in a floating-point operation.
  FILE_NOT_FOUND      Filename argument names a file that can't be found.
  FILE_NOT_READABLE   Filename argument is not a regular file.
  FILE_NOT_ACCESSIBLE Filename argument names a file that is read-protected.
  FILE_ERROR          Other error with filename argument.
  FILE_NOT_OPEN       A file function was given a channel that is not open.
  FILE_ERROR_-n       A file function got error -n ( Section 1.22).
  LOOKUP_FAILURE      Error looking up function (shouldn't happen).
  MALLOC_FAILURE      Failure to allocate needed memory (shouldn't happen).
  NAME_AMBIGUOUS      The function is not uniquely identified.
  MISSING_ARG         A required argument is missing.
  NO_SUCH_FUNCTION    An argument references a function that is not defined.
  NO_SUCH_MACRO       An argument references a macro that is not defined.
  RESULT_TOO_LONG     The result of a function is too long.
  UNKNOWN_FUNCTION    Internal error locating function (shouldn't happen).

   Examples:

  assign \%m \fmod()
  ?<ERROR:MISSING_ARG:\fmod()>
  echo "\fcontents(\%m)"
  "<ERROR:MISSING_ARG:\fmod()>"
  echo \fmod(3,x)
  ?<ERROR:ARG_NOT_NUMERIC:\fmod()>
  echo \fmod(3,4-2*2)
  ?<ERROR:DIVIDE_BY_ZERO:\fmod()>

   Notice the use of \fcontents() in echoing the value of a variable that
   contains a returned error message. That's because the error message
   includes the name of the variable or function that failed, so you must
   use \fcontents() to prevent it from being evaluated again -- otherwise
   the same error will occur.

   The handling of function and variable errors is controlled by:

   SET FUNCTION ERROR { ON, OFF }
          Tells whether invalid function calls or variable references
          should cause command errors. FUNCTION ERROR is ON by default.
          When ON, and an error is diagnosed in a built-in function or
          variable, the command that includes the function call or
          variable reference fails. The failing command can be handled in
          the normal way with IF FAILURE / IF SUCCESS, SET TAKE ERROR, or
          SET MACRO ERROR.

   When FUNCTION DIAGNOSTICS is OFF, there is no error message.

   SHOW SCRIPTS displays the current FUNCTION DIAGNOSTICS and ERROR
   settings.

  7.13. Return Value of Macros

   In C-Kermit 5A and 6.0, there are two ways to return one level from a
   macro: RETURN value and END number text. When RETURN is used, the
   value, which can be a number or a text string, is assigned to
   \v(return). When END was used, however, \v(return) was not set.
   SUCCESS/FAILURE was set according to whether the number was zero, and
   the text was printed, but the actual value of the number was lost.

   In C-Kermit 7.0, the END number is available in the \v(return)
   variable.

  7.14. The ASSERT, FAIL, and SUCCEED Commands.

   The ASSERT command is just like the IF command, but without a command
   to execute. It simply succeeds or fails, and this can be tested by a
   subsequent IF SUCCESS or IF FAILURE command. Example:

  ASSERT = 1 1
  IF SUCCESS echo 1 = 1.

   The FAIL command does nothing, but always fails. The SUCCEED command
   does nothing, but always succeeds.

   These commands are handy in debugging scripts when you want to induce a
   failure (or success) that normally would not occur, e.g. for testing
   blocks of code that normally are not executed.

  7.15. Using Alarms

   Alarms may be set in two ways:

   SET ALARM number
          Sets an alarm for the given number of seconds "from now", i.e.
          in the future, relative to when the SET ALARM command was given.
          Examples:

  set alarm 60        ; 60 seconds from now
  set alarm +60       ; The same as "60"
  set alarm -60       ; Not legal - you can't set an alarm in the past.
  set alarm 60*60     ; 60 minutes from now.
  set alarm \%a+10    ; You can use variables, etc.

   SET ALARM hh:mm:ss
          Sets an alarm for the specified time. If the given time is
          earlier than the current time, the alarm is set for the given
          time in the next day. You may give the time in various formats:

  set alarm 15:00:00  ; 3:00:00pm
  set alarm 3:00:00pm ; 3:00:00pm
  set alarm 3:00pm    ; 3:00:00pm
  set alarm 3pm       ; 3:00:00pm

   SHOW ALARM
          Displays the current alarm, if any, in standard date-time format
          (see Section 1.6): yyyymmdd hh:mm:ss.

   IF ALARM command
          Executes the command if an alarm has been set and the alarm time
          has passed.

   IF ALARM { command-list } [ ELSE { command-list } ]
          Executes the command-list if an alarm has been set and the alarm
          time has passed. Otherwise, if an ELSE part is given, its
          command-list is executed.

   CLEAR ALARM
          Clears the alarm.

   Only one alarm may be set at a time.

   Example: Suppose you have a script that is always running, and that
   transfers files periodically, and that keeps a transaction log. Suppose
   you want to start a new transaction log each day:

  log transactions \v(date).log
  set alarm 00:00:00                     ; Set an alarm for midnight
  while true {                           ; Main script loop
      xif alarm {                        ; If the alarm time is past...
          close transactions             ; Close current log
          log transactions \v(date).log  ; Start new one
          pause 1                        ; To make sure 00:00:00 is past
          set alarm 00:00:00             ; Set a new alarm
      }
      ; put the rest of the script here...
  }

   Note that IF ALARM -- no matter whether it succeeds or fails -- does
   NOT clear an expired alarm. Thus, once an alarm has expired, every IF
   ALARM will succeed until the alarm is cleared (with the CLEAR ALARM
   command) or reset with a new SET ALARM command.

  7.16. Passing Arguments to Command Files

   Beginning in version 7.0, C-Kermit accepts arguments on the TAKE
   command line, for example:

  C-Kermit> take oofa.ksc one two {this is three} four

   This automatically sets the variables \%1 through \%9 to the arguments,
   and \%0 to the name of the file, in this case:

  \%0 = /usr/olga/oofa.ksc
  \%1 = one
  \%2 = two
  \%3 = this is three
  \%4 = four

   and \%5..\%9 are undefined (empty). Arguments past the ninth are
   available in the \&_[] argument-vector array ( Section 7.5).

   The variables are those at the current macro level. Thus, if the TAKE
   command is executed from within a macro, the macro's arguments are
   replaced by those given on the TAKE command line (but only if at least
   one argument is given). The command shown above is exactly equivalent
   to:

  assign \%0 /usr/olga/oofa.ksc
  assign \%1 one
  assign \%2 two
  assign \%3 this is three
  assign \%4 four
  assign \%5
  assign \%6
  assign \%7
  assign \%8
  assign \%9
  take oofa.ksc

   Remember, the variables \%0..\%9 are on the macro call stack, and
   command files are independent of the macro stack. Thus, if a command
   file TAKEs another command file and passes arguments to it, the
   variables are changed from that point on for both files, and so forth
   for all levels of nested command files without intervening macro
   invocations.

   It would have been possible to change C-Kermit to use the overall
   command stack, rather than the macro stack, for arguments -- this would
   have made TAKE work exactly like DO, which is "nicer", but it would
   also have broken countless existing scripts. However, the new SHIFT
   command ( Section 7.5) makes it possible to create an alternative
   TAKE command that does indeed save and restore the argument variables
   at its own level around execution of a command file:

  define mtake {
     local \%f
     assign \%f \fcontents(\%1)
     shift
     take \%f
  }

   C-Kermit 7.0 also supports a new, easier way to pass arguments to
   scripts from the system command line:

  kermit filename arg1 arg2 arg3 ...

   in which arg1, arg2, arg3 (etc) are arguments for the script (whose
   filename is given), and are assigned to \%1, \%2, ... \%9. The filename
   is assigned to \%0. This applies equally to "Kerbang" scripts in UNIX
   ( Section 7.19). For example, suppose you have a file called
   "showargs" containing the following lines:

  #!/usr/local/bin/kermit +
  echo Hello from \%0
  show args
  exit

   (except not indented, since the "#!" line must be on the left margin).
   If you give this file execute permission:

  chmod +x showargs

   then you can run it exactly as you would run a UNIX shell script, e.g.:

  $ showargs one two three
  Hello from /usr/olga/showargs
  Top-level arguments (\v(argc) = 4):
   \&_[0] = /usr/olga/showargs
   \&_[1] = one
   \&_[2] = two
   \&_[3] = three

   Furthermore, the \&_[] array now contains the filename, if one was
   given as the first command line argument, or it is a "Kerbang" script,
   in element 0.

   Otherwise element 0 is program name, and elements 1 through \v(argc)-1
   contain the command-line arguments, if any, that appear after "--" or
   "=", if any. This array is saved and restored around macro calls;
   recall that inside macros it contains the macro argument vector
   (allowing you to access arguments programmatically, and to have more
   than 9 of them).

   At top level, notice the difference between the \&@[] and \&_[] arrays.
   The former includes C-Kermit options; the latter omits them.

  7.17. Dialogs with Timed Responses

   The ASK, ASKQ, GETOK, and GETC commands (let's call them the "ASK-class
   commands") let you write scripts that carry on dialogs with the user,
   asking them for text, a Yes/No answer, or a character, respectively.
   Prior to C-Kermit 7.0, these questions would always wait forever for an
   answer. In C-Kermit 7.0, you may specify a time limit for them with the
   new command:

   SET ASK-TIMER number
          Sets a time-limit on ASK-CLASS commands to the given number of
          seconds. If the number is 0 or less, there is no time limit and
          these commands wait forever for a response. Any timer that is
          established by this command remains in effect for all future
          ASK-class commands until another SET ASK-TIMER command is given
          (e.g. with a value of 0 to disable ASK timeouts).

   IF ASKTIMEOUT command
          An ASK-class command that times out returns a failure status.
          You can test explicitly for a timeout with:

  7.18. Increased Flexibility of SWITCH Case Labels

   Prior to C-Kermit 7.0 / K95 1.1.19, the case labels in SWITCH
   statements were string constants.

   Now case labels can be variables, function calls, or any mixture of
   these with each other and/or with regular characters.

   Furthermore, after the case label is evaluated, it is treated not as a
   string constant, but as a pattern against which the SWITCH variable is
   matched ( Section 4.9.1).

   This introduces a possible incompatibility with previous releases,
   since the following characters in case labels are no longer taken
   literally:

  \ * ? [ {

   Any scripts that previously included any of these characters in case
   labels must now quote them with backslash (\).

  7.19. "Kerbang" Scripts

   In UNIX only, Kermit scripts can be stored in files and run "directly",
   without starting Kermit first (as noted on page 467 of the manual),
   just as a shell script can be "run" as if it were a program. This
   section amplifies on that idea a bit, and presents some new aspects of
   version 7.0 that make it easier to write and run Kermit scripts
   directly.

     NOTE: On non-UNIX platforms, such as VMS or Windows, Kerbang scripts
     can be run as "kermit + scriptfilename arg1 arg2 arg3 ...". Windows
     95/98/NT file associations do not allow for the passing of
     parameters. In VMS, however, you can achieve the Kerbang effect by
     defining a symbol, as in this example:

  $ autotelnet :== "$SYS$TOOLS:KERMIT.EXE + AUTOTELNET.KSC"

     and then running the script like any other command:

  $ autotelnet xyzcorp.com myuserid

     See Section 9.3 for an explanation of the "+" symbol.

   UNIX shell scripts can specify which shell should run them by including
   a "shebang" line at the top, e.g.:

  #!/bin/sh

   (but not indented; the shebang line must be on the left margin). The
   term "shebang" is a contraction of "shell" and "bang". "Bang" is a
   slang word for the exclamation mark ("!"); "shebang" itself is an
   American slang word used in in the phrase "the whole shebang".

   We can run Kermit scripts directly too, by including a "shebang" line
   that names Kermit as the "shell"; thus we call these "Kerbang" scripts.
   This mechanism has been considerably simplified in C-Kermit 7.0 to
   facilitate C-Kermit's use a scripting tool just like any of the UNIX
   shells or scripting languages. The rules are the same as for shell
   scripts:

    1. The first line of the Kermit script must begin with "#!"
       immediately followed by the full pathname of the program that will
       execute the script (in this case, C-Kermit rather than a UNIX
       shell), followed by any Kermit command-line options. To suppress
       execution of the C-Kermit initialization file and to make command
       line arguments available to the script, the final option should be
       "+":
  #!/usr/local/bin/kermit +

       Some users have reported that in some circumstances a space might
       be necessary after the plus sign; this depends on your shell -- it
       has nothing to do with Kermit. In most cases, no space is needed.
    2. The file must have execute permission (granted via "chmod +x
       filename").

   When C-Kermit is invoked from a Kerbang script (or from the system
   prompt with a "+" command-line argument, which amounts to the same
   thing), the following special rules apply:

    1. The C-Kermit initialization file is NOT executed automatically. If
       you want it to be executed, include a TAKE command for it in the
       script, e.g. "take \v(home).kermrc". (In previous releases, the
       initialization file was always executed, with no way to prevent it
       except for the user to include Kermit-specific command line options
       which had nothing to do with the script). Many scripts have no need
       for the standard Kermit initialization file, which is quite lengthy
       and not only delays startup of the script, but also spews forth
       numerous messages that are most likely unrelated to the script.
    2. If the initialization file is not executed, neither is your
       customization file, since the initialization file is the command
       file from which the customization file is TAKEn. Again, you can
       include a TAKE command for the initialization file if desired, or
       for the customization file by itself, or for any other file.
    3. C-Kermit does not process command-line arguments at all. Instead,
       it passes all words on the command line after the "+" to the script
       as \%0 (the script name), \%1..\%9 (the first nine arguments), as
       well as in the argument vector array \&_[]. The variable \v(argc)
       is set to the total number of "words" (as passed by the shell to
       Kermit) including the script name. Quoting and grouping rules are
       those of the shell.
    4. At any point where the script terminates, it must include an EXIT
       command if you want it to exit back to the shell; otherwise
       C-Kermit enters interactive prompting mode when the script
       terminates. The EXIT command can include a numeric status to be
       returned to the shell (0, 1, etc), plus an optional message.

   Here is a simple Kerbang script that prints its arguments:

  #/usr/local/bin/kermit +
  echo Hello from \%0
  for \%i 0 \v(argc)-1 1 {
      echo \%i. "\&_[\%i]"
  }
  exit 0

   Save this file as (say) "showargs", then give it execute permission and
   run it (the \&_[] array is the same as \%0..\%9, but allows you to
   refer to argument variables programmatically; see Section 7.5).
   (Yes, you could substitute SHOW ARGUMENTS for the loop.)

  $ chmod +x showargs
  $ ./showargs one "this is two" three

   The script displays its arguments:

  Hello from /usr/olga/showargs
  0. "/usr/olga/showargs"
  1. "one"
  2. "this is two"
  3. "three"
  $

   Notice that no banners or greetings are printed and that startup is
   instantaneous, just like a shell script. Also notice that grouping of
   arguments is determined by *shell* quoting rules, not Kermit ones,
   since the command line is parsed by the shell before Kermit ever sees
   it.

   Of course you can put any commands at all into a Kerbang script. It can
   read and write files, make connections, transfer files, anything that
   Kermit can do -- because it *is* Kermit. And of course, Kerbang scripts
   can also be executed from the Kermit prompt (or from another script)
   with a TAKE command; the Kerbang line is ignored since it starts with
   "#", which is a comment introducer to Kermit just as it is to the UNIX
   shell. In VMS and other non-UNIX platforms, the Kerbang line has no
   effect and can be omitted.

   It might be desirable for a script to know whether it has been invoked
   directly from the shell (as a Kerbang script) or by a TAKE command
   given to the Kermit prompt or in a Kermit command file or macro. This
   can be done as in this example:

  #!/usr/local/bin/kermit +
  assign \%m \fbasename(\%0)
  define usage { exit 1 {usage: \%m phonenumber message} }
  define apage { (definition of APAGE...) } ; (Seebook pp.454-456)
  xif equal "\%0" "\v(cmdfil)" {
      if not def \%1 usage
      if not def \%2 usage
      apage {\%1} {\%2}
      exit \v(status)
  }

   In a Kerbang script, \%0 and \v(cmdfile) are the same; both of them are
   the name of the script. When a script is invoked by a Kermit TAKE
   command, \%0 is the name of the Kermit program, but \v(cmdfile) is the
   name of the script. In the example above, a macro called APAGE is
   defined. If the script was invoked directly, the APAGE macro is also
   executed. Otherwise, it is available for subsequent and perhaps
   repeated use later in the Kermit session.

   An especially handy use for Kerbang scripts is to have the
   initialization file itself be one. Since the standard initialization
   file is rather long and time-consuming to execute, it is often overkill
   if you want to start Kermit just to transfer a file. Of course there
   are command-line switches to suppress initialization-file execution,
   etc, but another approach is to "run" the initialization file when you
   want its features (notably the services directory), and run C-Kermit
   directly when you don't. A setup like this requires that (a) the
   C-Kermit initialization file is configured as a Kerbang script (has
   #!/path.../kermit as first line), has execute permission, and is in
   your PATH; and (b) that you don't have a .kermrc file in your login
   directory.

  7.20. IF and XIF Statement Syntax

   The IF command has been improved in two significant ways in C-Kermit
   7.0, described in the following subsections. All changes are backwards
   compatible.

    7.20.1. The IF/XIF Distinction

   The distinction between IF and XIF is no longer important as of
   C-Kermit 7.0. You should be able to use IF in all cases (and of course,
   also XIF for backwards compatibility). In the past, IF was used for
   single-command THEN parts, followed optionally by a separate ELSE
   command:

  IF condition command1    ; THEN part
  ELSE command2            ; ELSE part

   whereas XIF was required if either part had multiple commands:

  XIF condition { command, command, ... } ELSE { command, command, ... }

   The syntactic differences were primarily that IF / ELSE was two
   commands on two separate lines, whereas XIF was one command on one
   line, and that XIF allowed (and in fact required) braces around its
   command lists, whereas IF did not allow them.

   Furthermore, the chaining or nesting of parts and conditions was
   inconsistent. For example, the IF command could be used like this:

  IF condition command
  ELSE IF condition command
  ELSE IF condition command
  ELSE IF condition command
  ...

   but XIF could not. C-Kermit 7.0 accepts the old syntax and executes it
   the same as previous versions, but also accepts a new unified and more
   convenient syntax:

   IF condition command-list [ ELSE command-list ]

   or:

IF condition command-list
ELSE command-list

   in which the ELSE part is optional, and where command-list can be a
   single command (with or without braces around it) or a list of commands
   enclosed in braces. Examples:

   Example 1:

  IF condition { command1, command2 } ELSE { command3, command4 }

   Example 2 (same as Example 1):

  IF condition {
     command1
     command2
  } ELSE {
     command3
     command4
  }

   Example 3 (same as 1 and 2):

  IF condition {
     command1
     command2
  }
  ELSE { command3, command4 }

   Example 4 (same as 1-3):

  IF condition {
     command1
     command2
  }
  ELSE {
     command3
     command4
  }

   Example 5 (ELSE can be followed by another command):

  IF condition1 {
     command1
     command2
  } ELSE IF condition2 {
     command3
     command4
  } ELSE {
     command5
     command6
  }

   Example 5 suggests other possibilities:

  IF condition {
     command1
     command2
  } ELSE FOR variable initial final increment {
     command3
     command4
  }

   And this too is possible, except for some non-obvious quoting
   considerations:

  dcl \&a[6] = one two three four five six

  IF < \%n 3 {
      echo \\%n is too small: \%n
  } ELSE FOR \\%i 1 \\%n 1 {
      echo \\%i. \\&a[\\%i]
  }

   (The loop variable must be quoted in this context to prevent premature
   evaluation.)

   Many C programmers prefer to code IF-ELSE, WHILE, FOR, and SWITCH with
   the block-open bracket on its own line. This does not work in Kermit:

  IF condition        ; THIS FORMAT DOES NOT NOT WORK
  {
     command1
     command2
  }
  ELSE
  {
     command3
     command4
  }

   Explanation: the Kermit command language is line oriented; each line is
   a command, each command is a line. The first line above, having no hint
   of continuation, is an incomplete command, yet syntactically correct --
   an IF statement with an empty THEN part. Interestingly enough, since
   the next line begins with "{" it is a block that (inC-Kermit 8.0
   and later) is a block that is executed unconditionally. Thus the
   commands in the THEN part are executed regardless of whether the
   condition is true -- not what you wanted!

   The new block syntax used in the IF, WHILE, FOR, and SWITCH commands
   employs certain tricks to allow multiple lines to be treated as a
   single line:

     * Any line ending with "{" (ignoring whitespace and comments) marks
       the beginning of a block;
     * Any line beginning with "}" (ignoring whitespace) marks the end of
       a block;
     * Line breaks within a block separate commands; the comma is implied
       by the line end.

   Thus:

  IF condition {
     command1
     command2
  } ELSE {
     command3
     command4
  }

   is "assembled" into:

  IF condition { command1, command2 } ELSE { command3, command4 }

   Note the addition of commas to separate commands within blocks. As
   always, if you need continue a command onto additional lines, you can
   end the continued lines with the continuation character, "-". You can
   also do this if you want to put opening brackets on their own line:

  IF condition -
  {
     command1
     command2
  }
  ELSE -
  {
     command3
     command4
  }

    7.20.2. Boolean Expressions (The IF/WHILE Condition)

   Prior to C-Kermit 7.0, the IF and WHILE commands accepted only a single
   Boolean ("true or false") assertion, e.g. "if > \%m 0 command" or "if
   exist filename command". There was no way to form Boolean expressions
   and, in particular, nothing that approached a Boolean OR function (AND
   could be simulated by concatenating IF statements: "if condition1 if
   condition2..").

   C-Kermit 7.0 (and K95 1.1.19) allow grouping of Boolean assertions
   using parentheses and combining them using AND (or &&) and OR (or ||).
   Each of these operators -- including the parentheses -- is a field and
   must be set off by spaces. AND has higher precedence than OR, NOT has
   higher precedence than AND, but parentheses can be used to force any
   desired order of evaluation. The old syntax is still accepted.

   Here are some examples:

  define \%z 0                          ; Define some variables
  define \%n 1                          ; for use in the examples.

  if > \%n \%z echo \%n is greater.     ; Original format - still accepted.
  if ( > \%n \%z ) echo \%n is greater. ; Parentheses may be used in 7.0.
  if ( > \%n \%z && not = \%z 0 ) ...   ; Two assertions combined with AND.
  if ( > \%n \%z and not = \%z 0 ) ...  ; Same as previous ("and" = "&&").
  if ( > \%n \%z || not = \%z 0 ) ...   ; Two assertions combined with OR.
  if ( > \%n \%z or not = \%z 0 ) ...   ; Same as previous ("or" = "||").
  if ( > \%n \%z || != \%z 0 ) ...      ; Ditto ("!=" = "not =").
  while ( 1 ) { ... }                   ; Just like C.

   Notice the spaces around all operators including the parentheses --
   these are required. The following examples show how parentheses can be
   used to alter the precedence of the AND and OR operators:

  if ( false || false && false || true ) ,..         ; True
  if ( false || ( false && false ) || true ) ...     ; Same as previous
  if ( ( false || false ) && ( false || true ) ) ... ; False

   Similarly for NOT:

  if ( not true && false ) ...          ; False (NOT binds to TRUE only)
  if ( ( not true ) && false ) ...      ; Same as previous
  if ( not ( true && false ) ) ...      ; True (NOT binds to (TRUE && FALSE))

   Notes:

    1. The syntax of the Boolean expression itself has not changed; each
       expression begins with a keyword or token such as "EXIST", ">", or
       "=", etc; operators such as "<", "=", and ">" do not go between
       their operands but precede them as before; this might be called
       "reverse reverse Polish notation"; it allows deterministic
       on-the-fly parsing of these expressions at the C-Kermit> prompt as
       well as in scripts, and allows ?-help to be given for each item
       when IF or WHILE commands are typed at the prompt.
    2. Parentheses are required when there is more than one Boolean
       assertion.
    3. Parentheses are not required, but are allowed, when there is only
       one Boolean assertion.
    4. Evaluation of Boolean assertions occurs left to right, but the
       resulting Boolean expression is evaluated afterwards according to
       the rules of precedence. All Boolean assertions are always
       evaluated; there is no "early stopping" property and therefore no
       question about when or if side effects will occur -- if any Boolean
       assertion has side effects, they will always occur. (Early stopping
       is, however, possible with theS-Expression IF introduced in
       C-Kermit 8.0.)

   Constructions of arbitrary complexity are possible, within reason.

   Also see Section 7.4 for new IF / WHILE conditions.

  7.21. Screen Formatting and Cursor Control

   C-Kermit 7.0 adds a simple way to create formatted screens, the SCREEN
   command:

   SCREEN { CLEAR, CLEOL, MOVE-TO row [ column ] }
          Performs screen-formatting actions. Correct operation of these
          commands depends on proper terminal setup on both ends of the
          connection -- mainly that the host terminal type is set to agree
          with the kind of terminal or the emulation you are viewing
          C-Kermit through. The UNIX version uses terminfo or termcap (not
          curses); the VMS version uses SMG; K-95 uses its built in screen
          manager.

   SCREEN CLEAR
          Moves the cursor to home position and clears the entire screen.
          Synonyms: CLEAR COMMAND-SCREEN ALL (K-95 only), CLS, CLEAR
          SCREEN.

   SCREEN CLEOL
          Clears from the current cursor position to the end of the line.
          Synonym: CLEAR COMMAND-SCREEN EOL (K-95 only)

   SCREEN MOVE-TO row column
          Moves the cursor to the indicated row and column. The row and
          column numbers are 1-based, so on a 24x80 screen the home
          position is 1 1 and the lower right corner is 24 80. If a row or
          column number is given that too large for what Kermit or the
          operating system thinks is your screen size, the appropriate
          number is substituted.

   These escape sequences used by these commands depends on the platform.
   In UNIX, your TERM environment variable is used to query the
   terminfo/termcap database; if the query fails, ANSI/VT100 sequences are
   used. In VMS, the SMG library is used, which sends sequences based on
   your VMS terminal type. K95 does its own screen control. On other
   platforms (such as AOS/VS, VOS, etc), screen formatting is not
   supported, and the SCREEN command does nothing.

   The three SCREEN actions can be used in scripts to produce menus,
   formatted screens, dynamic displays, etc. Related variables include:

  \v(terminal)     The type terminal C-Kermit thinks you have.
  \v(rows)         The number of rows C-Kermit thinks your terminal has.
  \v(columns)      The number of columns C-Kermit thinks your terminal has.

   And functions:

  \fscrncurx()     The current X coordinate of the cursor (K-95 only).
  \fscrncury()     The current Y coordinate of the cursor (K-95 only).
  \fscrnstr(x,y,n) The string of length nat position (x,y) (K-95 only).

   And commands:

  ECHO string      Writes string + CRLF at the current cursor position.
  XECHO string     Writes string at current cursor position; CRLF not supplied.
  GETC v prompt    Issues prompt, reads one character into variable v, no echo.

   And special characters:

  Ctrl-L           At the C-Kermit> command prompt, or in a C-Kermit command,
                   works like Return or Enter, but also clears the screen

   Example 1: A macro that prints a message \%1 at cursor position
   (\%2,\%3):

  define MSG {
      if not def \%3 def \%3 0             ; Default column to 0
      if > \v(argc) 2 screen move \%2 \%3  ; Move to given row/col (if any)
      screen cleol                         ; Clear to end of line
      if def \%1 xecho \fcontents(\%1)     ; Print message (if any)
  }

   Example 2: A macro put the cursor on the bottom screen line, left
   margin:

  define BOT {
      screen move \v(rows) 0
  }

   Example 3: A macro to center message \%1 on line \%2.

  define CENTER {
      if not def \%2 def \%2 1
      .\%x ::= (\v(cols)-\flen(\%1))/2
      msg {\%1} {\%2} {\%x}
  }

   Example 4: A simple menu (building on Examples 1-3):

  def \%c 0                             ; Menu choice variable
  screen clear                          ; Clear the screen
  center {Welcome to This Menu} 2       ; Display the menu
  msg {Choices:} 4
  msg { 1. File} 6
  msg { 2. Edit} 7
  msg { 3. Exit} 8
  while ( != \%c 3 ) {                  ; Read and verify choice
      while true {                      ; Keep trying till we get a good one
          screen move 10                ; Move to line 10
          screen cleol                  ; Clear this line
          getc \%c {Your choice: }      ; Prompt and get and echo 1 character
          xecho \%c
          if ( not numeric \%c ) { msg {Not numeric - "\%c"} 12, continue }
          if ( >= \%c 1 && <= \%c 3 ) break
          msg {Out of range - "\%c"} 12
      }
      switch \%c {                      ; Valid choice - execute it.
        :1, msg {Filing... } 12, break
        :2, msg {Editing...} 12, break
        :3, msg {Exiting...} 12, break
      }
  }
  echo Bye                              ; Exit chosen - say goodbye.
  bot                                   ; Leave cursor at screen bottom.
  exit                                  ; And exit.

   Similar scripts can work over the communication connection; substitute
   INPUT and OUTPUT for GETC and ECHO/XECHO.

  7.22. Evaluating Arithmetic Expressions

   A new arithmetic operator was added to the list recognized by the
   EVALUATE command, the \feval() function, and which can also be used
   anywhere else arithmetic expressions are accepted (numeric command
   fields, array subscripts, etc):

   Prefix "!"
          This operator inverts the "truth value" of the number or
          arithmetic expression that follows. If the value of the operand
          is 0, the result is 1. If the value is nonzero, the result is 0.

   Examples:

  set eval old
  evaluate 0
  0

  evaluate !0
  1

  evaluate !3
  0

  evaluate !(-3)
  0

  .\%a = 1
  .\%b = 0
  evaluate !(\%a|\%b)
  0

  evaluate !(\%a&\%b)
  1

  evaluate !(!(\%a&\%b))
  0

   Note the distinction between Prefix ! (invert truth value) and Suffix !
   (factorial). Also the distinction between Prefix ! and Prefix ~ (which
   inverts all the bits in its operand). Also note that prefix operators
   (!, -, and ~) can not be adjacent unless you use parentheses to
   separate them, as shown in the final example above.

  7.23. Floating-Point Arithmetic

     For a more convenient way of dealing with floating-point numbers
     than the one described here, see theC-Kermit 8.0 update notes,
     the section onS-Expressions.

   C-Kermit 7.0 adds limited support for floating-point numbers (numbers
   that have fractional parts, like 3.141592653). This support is provided
   through a small repertoire of functions and in Boolean expressions that
   compare numbers, but does not apply to number parsing in general, or to
   expression evaluation, array subscripts, the INCREMENT and DECREMENT
   commands, or in any context other than those listed in this section.

   A floating point number has an optional sign (+ or -), followed by a
   series of decimal digits containing either zero or one period (.)
   character, which is the decimal point. The use of comma or any other
   character besides period as a decimal point is not supported.
   Scientific notation is not supported either. Examples of legal
   floating-point numbers:

  0                Integers can be used
  1                Ditto
  2.               A decimal point without decimal digits
  3.0              A decimal point with decimal digits
  3.141592653      Ditto
 -4.0              A negative sign can be included
 +5.0              A positive sign can be included

   Examples of notations that are not accepted:

  1,000,000        Separators can not be used
  1.000.000        Ditto (or multiple decimal points)
  6.022137E23      No scientific notation
  6.62606868e-34   Ditto
  12.5+6.25        No "bare" expressions

   You can use IF FLOAT test a string or variable to see if it's in
   acceptable floating-point format. Example:

  ask \%f { Type a number: }
  if not def \%f .\%f = 0.0
  if not float \%f stop 1 Invalid floating-point number: "\%f"

   C-Kermit's floating-point support, like its support for whole numbers
   (integers), relies on the capabilities of the underlying computer. Your
   computer has only a limited amount of precision for numbers, depending
   on its architecture. Thus floating-point numbers that have too many
   digits will not be accurate; adding a very small number to a very large
   one might have no effect at all; and so on. For details, read a text on
   numerical analysis. Example:

  .\%a = 11111111111111111111  ; A long number
  .\%b = 22222222222222222222  ; Another one
  echo \ffpadd(\%a,\%b)        ; Add them - the result should be all 3's
  33333333333333330000.0       ; See the result

   In this example, the computer has 16 digits of precision; after that,
   the (low-order) digits are set to 0, since the computer doesn't know
   what they really are. In fact, the computer returns random digits, but
   Kermit sets all digits beyond the computer's precision to 0.

   C-Kermit's floating-point functions have names of the form
   "\ffpxxx(args)" ("\f" for function, "fp" for floating-point), where
   "xxx" is replaced by the name of the function, such as "sqrt", and
   "args" is the argument list, consisting of one or two floating-point
   numbers (depending on the function), and an optional "d" argument that
   says now many decimal places should be shown in the result. Example:

  \ffpdiv(10,3,1) returns "3.3"
  \ffpdiv(10,3,2) returns "3.33"
  \ffpdiv(10,3,3) returns "3.333"

   and so on, up to the precision of the computer. If the decimal-places
   argument is less than zero, the fractional part of the result is
   truncated:

  \ffpdiv(10,3,-1) returns "3".

   If the decimal-places argument is 0, or is omitted, C-Kermit returns as
   many decimal places as are meaningful in the computer's floating-point
   precision, truncating any extraneous trailing 0's:

  \ffpdiv(10,8) returns "1.25".
  \ffpdiv(10,4) returns "2.5".
  \ffpdiv(10,2) returns "5.0".
  \ffpdiv(10,3) returns "3.333333333333333" (for 16-digit precision).

   There is no way to request that a floating-point function return a
   decimal point but no decimal places. However, this is easy enough to
   accomplish in other ways, for example by supplying it outside the
   function call:

  echo \ffpadd(\%a,\%b,-1).

   Kermit's floating-point functions always round the result for the
   requested number of decimal places when the "d" argument is given and
   has a value greater than 0 (see the description of \ffpround() just
   below).

   Floating-point arguments can be constants in floating-point format or
   variables whose values are floating-point numbers. If a floating-point
   argument is omitted, or is a variable with no value, 0.0 is supplied
   automatically. Example:

  def \%x 999.999
  undef \%y
  echo \ffpmin(\%x,\%y)
  0.0

   Or equivalently:

  echo \ffpmin(999.999)
  0.0

   The floating-point functions are:

   \ffpround(f1,d)
          Returns f1 rounded to d decimal places. For this function only,
          d = 0 (or d omitted) has a special meaning: return the integer
          part of f1 rounded according to the fractional part. Examples:

  \ffpround(2.74653,-1) returns "2" (fraction truncated, no rounding).
  \ffpround(2.74653,0)  returns "3" (integer part is rounded).
  \ffpround(2.74653)    returns "3" (d omitted same as d = 0).
  \ffpround(2.74653,1)  returns "2.7".
  \ffpround(2.74653,2)  returns "2.75".
  \ffpround(2.74653,3)  returns "2.747".
  \ffpround(2.74653,4)  returns "2.7465", etc.

   \ffpadd(f1,f2,d)
          Returns the sum of f1 and f2.

   \ffpsubtract(f1,f2,d)
          Subtracts f2 from f1 and returns the result.

   \ffpmultiply(f1,f2,d)
          Returns the product of f1 and f2.

   \ffpdivide(f1,f2,d)
          If f2 is not 0, divides f1 by f2 and returns the quotient.
          If f2 is 0, a DIVIDE_BY_ZERO error occurs.

   \ffpraise(f1,f2,d)
          If f1 = 0 and f2 <= 0, or if f1 < 0 and f2 has a fractional
          part, an ARG_OUT_OF_RANGE error occurs; otherwise f1 raised to
          the f2 power is returned.

   \ffpsqrt(f1,d)
          If f1 >= 0, returns the square root of f1; otherwise
          ARG_OUT_OF_RANGE.

   \ffpabsolute(f1,d)
          Returns the absolute value of f1 (i.e. f1 without a sign). This
          is the floating-point analog of \fabsolute(n1).

   \ffpint(f1)
          Returns the integer part of f1. Equivalent to \ffpround(f1,-1).

   \ffpexp(f1,d)
          The base of natural logarithms, e (2.718282...), raised to the
          f1 power.

   \ffplogn(f1,d)
          The natural logarithm of f1 (the power to which e must be raised
          to obtain f1).

   \ffplog10(f1,d)
          The base-10 logarithm of f1 (the power to which 10 must be
          raised to obtain f1).

   \ffpmodulus(f1,f2,d)
          If f2 is not 0, the remainder after dividing f1 by f2.
          If f2 is 0, a DIVIDE_BY_ZERO error occurs.
          This is the floating-point analog of \fmod(n1,n2).

   \ffpmaximum(f1,f2,d)
          Returns the maximum of f1 and f2. This is the floating-point
          analog of \fmax(n1,n2).

   \ffpminimum(f1,f2,d)
          Returns the minimum of f1 and f2. This is the floating-point
          analog of \fmin(n1,n2).

   \ffpsine(f1,d)
          Returns the sine of f1 radians.

   \ffpcosine(f1,d)
          Returns the cosine of f1 radians.

   \ffptangent(f1,d)
          Returns the tangent of f1 radians.

   Note that all of these functions can be used with integer arguments. If
   you want an integer result, specify d = -1 (to truncate) or feed the
   result to \ffpround(xxx,0) (to round).

   Floating-point numbers (or variables or functions that return them) can
   be used in Boolean expressions (see Section 7.20.2) that compare
   numbers:

  = x y
  != x y
  < x y
  > x y
  <= x y
  >= x y

   In these examples, x and y can be either integers or floating-point
   numbers in any combination. In an arithmetic comparison of an integer
   and a floating-point number, the integer is converted to floating-point
   before the comparison is made. Examples:

  .\%t = 3.000000000
  .\%f = 3.141592653
  .\%i = 3

  if > \%f \%i echo Pi is greater.
  if = \%t \%i echo "\%i" = "\%t".

   A floating-point number can also be used in:

  IF number command

   where the command is executed if the number is nonzero. If the number
   is floating-point, the command is not executed if the number is 0.0,
   and is executed otherwise.

   Floating-point numbers can be sorted using ARRAY SORT /NUMERIC (see
  Section 7.10.5 ).

   Two floating-point constants are provided:

  \v(math_pi) = Pi (3.141592653...)
  \v(math_e)  = e, the base of natural logarithms (2.71828...)

   These are given to the computer's precision, e.g. 16 digits. This
   number itself is available in a variable:

   \v(math_precision)
          How many significant digits in a floating-point number.

  7.24. Tracing Script Execution

   The TRACE command is handy for debugging scripts.

   TRACE [ { /ON, /OFF } ] [ { ASSIGNMENTS, COMMAND-LEVEL, ALL } ]
          Selects tracing of the given object.

   Optional switches are /ON and /OFF. If no switch is given, /ON is
   implied. The trace objects are ASSIGNMENTS, COMMAND-LEVEL, and ALL. The
   default object is ALL, meaning to select all trace objects (besides
   ALL). Thus TRACE by itself selects tracing of everything, as does TRACE
   /ON, and TRACE /OFF turns off all tracing.

   When tracing of ASSIGNMENTS is on, every time the value of any
   user-defined variable or macro changes, C-Kermit prints one of the
   following:

   >>> name: "value"
          The name of the variable or macro followed by the new value in
          quotes. This includes implicit macro-parameter assignments
          during macro invocation.

   >>> name: (undef)
          This indicates that the variable or macro has been undefined.

   <<< name: "value"
          For RETURN statements: the name of the macro and the return
          value.

   <<< name: (null)
          For RETURN statements that include no value or an empty value.

   When tracing of COMMAND-LEVEL is on, C-Kermit prints:

   [n] +F: "name"
          Whenever a command file is entered, where "n" is the command
          level (0 = top); the name of the command file is shown in
          quotes.

   [n] +M: "name"
          Whenever a macro is entered; "n" is the command level. The name
          of the macro is shown in quotes.

   [n] -F: "name"
          Whenever a command file is reentered from below, when a macro or
          command file that it has invoked has returned.

   [n] -M: "name"
          Whenever a macro is reentered from below.

   For other debugging tools, see SHOW ARGS, SHOW STACK, SET TAKE, SET
   MACRO, and of course, ECHO.

  7.25. Compact Substring Notation

   It is often desirable to extract a substring from a string which is
   stored in a variable, and for this we have the \fsubstring() function,
   which is used like this:

  define \%a 1234567890
  echo \fsubstring(\%a,3,4) ; substring from 3rd character length 4
  3456

   or like this with macro-named variables:

  define string 1234567890
  echo \fsubstring(\m(string),3,4)
  3456

   C-Kermit 7.0 adds a pair of alternative compact notations:

\:(variablename[start:length])  <-- Substring of variable's value
\s(macroname[start:length])     <-- Substring of macro's definition

   These are exactly equivalent to using \fsubstring(), except more
   compact to write and also faster since evaluation is in one step
   instead of two.

   The "\:()" notation can be used with any Kermit variable, that is,
   almost anything that starts with a backslash:

  \:(\%a[2:6])      <-- equivalent to \fsubstring(\%a,2,6)
  \:(\&x[1][2:6])   <-- equivalent to \fsubstring(\&x[1],2,6)
  \:(\m(foo)[2:6])  <-- equivalent to \fsubstring(\m(foo),2,6)
  \:(\v(time)[2:6]) <-- equivalent to \fsubstring(\v(time),2,6)
  \:(\$(TERM)[2:6]) <-- equivalent to \fsubstring(\$(TERM),2,6)
  \:(ABCDEFGH[2:6]) <-- equivalent to \fsubstring(ABCDEFGH,2,6)

   Whatever appears between the left parenthesis and the left bracket is
   evaluated and then the indicated substring of the result is returned.

   The "\s()" notation is the same, except after evaluating the variable,
   the result is treated as a macro name and is looked up in the macro
   table. Then the indicated substring of the macro definition is
   returned. Example:

  define testing abcdefghijklmnopqrstuvwxyz
  define \%a testing

  \s(testing[2:6])  -->  bcdefg
  \:(testing[2:6])  -->  esting
  \:(\%a[2:6])      -->  esting
  \s(\%a[2:6])      -->  bcdefg

   Note that the following two examples are equivalent:

  \:(\m(foo)[2:6])
  \s(foo[2:6])

   The first number in the brackets is the 1-based starting position. If
   it is omitted, or less than 1, it is treated as 1. If it is greater
   than the length of the string, an empty string is returned.

   The second number is the length of the desired substring. If the second
   number is omitted, is less than 0, or would be past the end of the
   string, then "through the end of the string" is assumed. If it is 0,
   the empty string is returned.

   If the brackets are empty or omitted, the original string is returned.

   The starting position and length need not be literal numbers; they can
   also be variables, functions, arithmetic expressions, or even other
   \s() or \:() quantities; anything that evaluates to a number, for
   example:

  \s(block[1025:\fhex2n(\s(block[\%b:\%n+4]))/2])

   Syntactically, \m(name) and \s(name) differ only in that the sequence
   [*] at the end of the name (where * is any sequence of 0 or more
   characters) is treated as substring notation in \s(name), but is
   considered part of the name in \m(name) (to see why, seeSection
   7.10.9).

  7.26. New WAIT Command Options

   The WAIT command has been extended to allow waiting for different kinds
   of things (formerly it only waited for modem signals). Now it also can
   wait for file events.

    7.26.1. Waiting for Modem Signals

   The previous syntax:

  WAIT time { CD, DSR, RTS, RI, ... }

   has changed to:

  WAIT time MODEM-SIGNALS { CD, DSR, RTS, RI, ... }

   However, the previous syntax is still accepted. The behavior is the
   same in either case.

    7.26.2. Waiting for File Events

   The new WAIT option:

  WAIT time FILE { CREATION, DELETION, MODIFICATION } filename

   lets you tell Kermit to wait the given amount of time (or until the
   given time of day) for a file whose name is filename to be created,
   deleted, or modified, respectively. The filename may not contain
   wildcards. If the specified event does not occur within the time limit,
   or if WAIT CANCELLATION is ON and you interrupt from the keyboard
   before the time is up, the WAIT command fails. If the event is
   MODIFICATION and the file does not exist, the command fails. Otherwise,
   if the given event occurs within the time limit, the command succeeds.
   Examples:

   WAIT 600 FILE DELETION oofa.tmp
          Wait up to 10 minutes for file oofa.tmp to disappear.

   WAIT 23:59:59 FILE MOD orders.db
          Wait until just before midnight for the orders.db file to be
          changed.

   Example: Suppose you want to have the current copy of /etc/motd on your
   screen at all times, and you want to hear a bell whenever it changes:

  def \%f /etc/motd                      ; The file of interest.
  while 1 {                              ; Loop forever...
      cls                                ; Clear the screen.
      echo \%f: \v(date) \v(time)...     ; Print 2-line heading...
      echo
      if ( not exist \%f ) {             ; If file doesn't exist,
          echo \%f does not exist...     ; print message,
          wait 600 file creat \%f        ; and wait for it to appear.
          continue
      }
      beep                               ; Something new - beep.
      type /head:\v(rows-2) \%f          ; Display the file
      if fail exit 1 \%f: \ferrstring()  ; (checking for errors).
      wait 999 file mod \%f              ; Wait for it to change.
  }

   This notices when the file is created, deleted, or modified, and acts
   only then (or when you interrupt it with); the time shown in the
   heading is the time of the most recent event (including when the
   program started).

   See Section 1.10, where the \v(kbchar) variable is explained. This
   lets you modify a loop like the one above to also accept
   single-character commands, which interrupt the WAIT, and dispatch
   accordingly. For example:

  wait 999 file mod \%f              ; Wait for the file to change.
  if defined \v(kbchar) {            ; Interrupted from keyboard?
      switch \v(kbchar) {            ; Handle the keystroke...
        :q, exit                     ; Q to Quit
        :h, echo blah blah, break    ; H for Help
        :default, beep, continue     ; Anything else beep and ignore
      }
  }

   This lets you write event-driven applications that wait for up to three
   events at once: a file or modem event, a timeout, and a keystroke.

  7.27. Relaxed FOR and SWITCH Syntax

   For consistency with the extended IF and WHILE syntax, the FOR and
   SWITCH control lists may (but need not be) enclosed in parentheses:

  FOR ( \%i 1 \%n 1 ) { command-list... }
  SWITCH ( \%c ) { command-list... }

   In the FOR command, the increment item can be omitted if the control
   list is enclosed in parentheses, in which case the increment defaults
   appropriately to 1 or -1, depending on the values of the first two
   variables.

   As with IF, the parentheses around the FOR-command control list must be
   set off by spaces (in the SWITCH command, the spaces are not required
   since the SWITCH expression is a single arithmetic expression).

   Also, outer braces around the command list are supplied automatically
   if you omit them, e.g.:

  FOR ( \%i 1 %n 1 ) echo \%i

  8. USING OTHER FILE TRANSFER PROTOCOLS

   In C-Kermit 7.0, alternative protocols can be selected using switches.
   Switches are described in Section 1.5; the use of
   protocol-selection switches is described in Section 4.7.1.
   Example:

  send /binary /protocol:zmodem x.tar.gz

   Note that file transfer recovery works only with Kermit and Zmodem
   protocols. With Zmodem, recovery can be initiated only by the sender.

   Only pre-1988 versions of the publicly-distributed sz/rz programs use
   Standard I/O; those released later than that do not use Standard I/O
   and therefore do not work with REDIRECT. However, Omen Technology does
   offer an up-to-date redirectable version called crzsz, which must be
   licensed for use:

     "Unix Crz and Csz support XMODEM, YMODEM, and ZMODEM transfers when
     called by dial-out programs such as Kermit and certain versions of
     cu(1). They are clients designed for this use.

     "Crz and Csz are Copyrighted shareware programs. Use of these
     programs beyond a brief evaluation period requires registration.
     Please print the "mailer.rz" file, fill out the form and return same
     with your registration."

   To use the crzsz programs as your external XYZMODEM programs in
   C-Kermit, follow the instructions in the book, but put a "c" before
   each command, e.g.:

  set protocol zmodem {csz %s} {csz -a %s} crz crz crz crz

   To use Zmodem protocol over Telnet or other non-transparent
   connections, you might need to add the -e (Escape) option:

  set protocol zmodem {csz -e %s} {csz -e -a %s} crz crz crz crz

  9. COMMAND-LINE OPTIONS

  9.0. Extended-Format Command-Line Options

   Standard UNIX command line options are a single letter. C-Kermit has
   run out of letters, so new options are in a new extended format:

 --word[:arg]

   where a keyword (rather than a single letter) specifies the function,
   and if an argument is to be included, it is separated by a colon (or
   equal sign). Most of the new extended-format command-line options are
   only for use with the Internet Kermit Service Daemon; see theIKSD
   Administration Guide for details. However, several of them are also
   general in nature:

   --nointerrupts
          Disables keyboard interrupts that are normally enabled, which
          are usually Ctrl-C (to interrupt a command) and Ctrl-Z (UNIX
          only, to suspend C-Kermit).

   --help
          Lists the extended command-line options that are available in
          your version of C-Kermit. If any options seem to be missing,
          that is because your copy of C-Kermit was built with
          compile-time options to deselect them.

   --helpfile:filename
          Specifies the name of a file to be displayed if the user types
          HELP (not followed by a specific command or topic), in place of
          the built-in top-level help text. The file need not fit on one
          screen; more-prompting is used if the file is more than one
          screen long if COMMAND MORE-PROMPTING is ON, as it is by
          default.

   --bannerfile:filename
          The name of a file containing a message to be printed after the
          user logs in, in place of the normal message (Copyright notice,
          "Type HELP or ? for help", "Default transfer mode is...", etc).

   --cdmessage:{on,off,0,1,2}
          For use in the Server-Side Server configuration; whenever the
          client tells the server to change directory, the server sends
          the contents of a "read me" file to the client's screen. This
          feature is On by default, and operates only in client/server
          mode when ON or 1. If set to 2 or higher, it also operates when
          the CD command is given at the IKSD> prompt. Synonym: --cdmsg.

   --cdfile:filename
          When cdmessage is on, this is the name of the "read me" file to
          be sent. Normally you would specify a relative (not absolute)
          name, since the file is opened using the literal name you
          specified, after changing to the new directory. Example:

  --cdfile:READ.ME

          You can also give a list of up to 8 filenames by (a) enclosing
          each filename in braces, and (b) enclosing the entire list in
          braces. Example:
          --cdfile:{{./.readme}{READ.ME}{aaareadme.txt}{README}{read-this-
          first}} When a list is given, it is searched from left to right
          and the first file found is displayed. The default list for UNIX
          is:

  {{./.readme}{README.TXT}{READ.ME}}

  9.1. Command Line Personalities

   Beginning in version 7.0, if the C-Kermit binary is renamed to "telnet"
   (or TELNET.EXE, telnet.pr, etc, depending on the platform), it accepts
   the Telnet command line:

  telnet [ host [ port ] ]

   In Unix, you can achieve the same effect with a symlink:

  cd /usr/bin
  mv telnet oldtelnet
  ln -ls /usr/local/bin/kermit telnet

   When installed in this manner, C-Kermit always reads its initialization
   file. If no host (and therefore no port) is given, C-Kermit starts in
   interactive prompting mode. If a host is given as the first
   command-line argument, C-Kermit makes a connection to it. The host
   argument can be an IP host name or address, or the name of a TCP/IP
   entry in your C-Kermit network directory.

   If a port is given, it is used. If a port is not given, then if the
   hostname was found in your network directory and port was also listed
   there, then that port is used. Otherwise port 23 (the Telnet port) is
   used.

   When C-Kermit is called "telnet" and it is invoked with a hostname on
   the command line, it exits automatically when the connection is closed.
   While the connection is open, however, you may escape back and forth as
   many times as you like, transfer files, etc.

   An rlogin personality is also available, but it is less useful, at
   least in UNIX and VMS, where the Rlogin TCP port is privileged.

   The new variable \v(name) indicates the name with which C-Kermit was
   invoked ("kermit", "wermit", "k95", "telnet", etc).

  9.2. Built-in Help for Command Line Options

   "kermit -h", given from the system prompt, lists as many command-line
   options as will fit on a standard 24x80 screen. For more comprehensive
   help, use the interactive HELP OPTIONS command that was added in
   C-Kermit 7.0:

   HELP OPTIONS
   Explains how command-line options work, their syntax, etc.

   HELP OPTIONS ALL
   Lists all command-line options and gives brief help about each one.

   HELP OPTION x
   Gives brief help about option "x".

   HELP EXTENDED-OPTIONS
   Lists the available extended-format command-line options.

   HELP EXTENDED-OPTION xxx
   Gives help for the specified extended option.

  9.3. New Command-Line Options

   Command-line options added since C-Kermit 6.0 are:

   +
          (plus sign by itself): The next argument is the name of a script
          to execute; all subsequent arguments are ignored by C-Kermit
          itself, but passed to the script as top-level copies of \%1,
          \%2, etc; the \&_[] is also set accordingly. \%0 and \&_[0]
          become the name of the script file, rather than the pathname of
          the C-Kermit program, which is its normal value. Primarily for
          use in the top line of "Kerbang" scripts in UNIX (see
         Section 7.19). Example from UNIX command line:

  $ kermit [ regular kermit args ] + filename

          Sample first line of Kerbang script:

  #!/usr/local/bin/kermit +

   --
          (two hyphens surrounded by whitespace) Equivalent to "=", for
          compatibility with UNIX getopt(1,3).

   -G
          GET (like -g), but send the incoming file to standard output.
          Example: "kermit -G oofa.txt | lpr" retrieves a file from your
          local computer (providing it is running a Kermit program that
          supports the autodownload feature and has it enabled) and prints
          it.

   -O
          equivalent to -x (start up in server mode), but exits after the
          first client command has been executed (mnemonic: O = Only One).
          This one is handy replacing "kermit -x" in the "automatically
          start Kermit on the other end" string:

  set protocol kermit {kermit -ir} {kermit -r} {kermit -x}

          since -x leaves the remote Kermit in server mode after the
          transfer, which can be confusing, whereas -O makes it go away
          automatically after the transfer.

   -L
          Recursive, when used in combination with -s (mnemonic: L =
          Levels). In UNIX or other environments where the shell expands
          wildcards itself, the -s argument, if it contains wildcards,
          must be quoted to prevent this, e.g.:

  kermit -L -s "*.c"

          In UNIX only, "kermit -L -s ." means to send the current
          directory tree. SeeSections 4.10 and4.11 about
          recursive file transfer.

   -V
          Equivalent to SET FILE PATTERNS OFF ( Section 4.3) and SET
          TRANSFER MODE MANUAL. In other words, take the FILE TYPE setting
          literally. For example, "kermit -VT oofa.bin" means send the
          file in Text mode, no matter what its name is and no matter
          whether a kindred spirit is recognized at the other end of the
          connection.

   -0
          (digit zero) means "be 100% transparent in CONNECT mode". This
          is equivalent to the following series of commands: SET PARITY
          NONE, SET COMMAND BYTESIZE 8, SET TERMINAL BYTESIZE 8, SET FLOW
          NONE, SET TERM ESCAPE DISABLED, SET TERM CHAR TRANSPARENT, SET
          TERM AUTODOWNLOAD OFF, SET TERM APC OFF, SET TELOPT KERMIT
          REFUSE REFUSE.

  10. C-KERMIT AND G-KERMIT

   Every multifunctioned and long-lived software program grows in
   complexity and size over time to meet the needs and requests of its
   users and the demands of the underlying technology as it changes.

   Eventually users begin to notice how big the application has grown, how
   much disk space it occupies, how long it takes to load, and they start
   to long for the good old days when it was lean and mean. Not long after
   that they begin asking for a "light" version that only does the basics
   with no frills.

   And so it is with C-Kermit. A "light" version of Kermit was released
   (for UNIX only) in December 1999 under the GNU General Public License;
   thus it is called G-Kermit (for GNU Kermit). All it does is send and
   receive files, period. You can find it at:

 http://www.columbia.edu/kermit/gkermit.html

   Where the C-Kermit 7.0 binary might be anywhere from 1 to 3 million
   bytes in size, the G-Kermit binary ranges from 30K to 100K, depending
   on the underlying architecture (RISC vs CISC, etc).

   G-Kermit and C-Kermit may reside side-by-side on the same computer.
   G-Kermit does not make connections; it does not have a script language;
   it does not translate character sets. G-Kermit may be used instead of
   C-Kermit when:

     * It is on the remote end.
     * Files are to be transferred in binary mode or in text mode without
       character-set translation.
     * File timestamps don't need to be preserved.

   In such cases G-Kermit might be preferred since it generally starts up
   faster, and yet transfers files just as fast on most (but not
   necessarily all) kinds of connections; for example, it supports
   streaming ( Section 4.20).

   G-Kermit is also handy for bootstrapping. It is easier to load on a new
   computer than C-Kermit -- it fits on a floppy diskette with plenty of
   room to spare. Thus if you have (say) an old PC running (say) SCO Xenix
   and no network connection, you can download the Xenix version of
   G-Kermit to (say) a DOS or Windows PC, copy it to diskette, read the
   diskette on Xenix with "dosread", and then use G-Kermit to receive
   C-Kermit (which does not fit on a diskette). If diskettes aren't an
   option, other bootstrapping methods are possible too -- see the
  G-Kermit web page for details.

III. APPENDICES

  III.1. Character Set Tables

    III.1.1. The Hewlett Packard Roman8 Character Set

dec col/row oct hex  description
160  10/00  240  A0  (Undefined)
161  10/01  241  A1  A grave
162  10/02  242  A2  A circumflex
163  10/03  243  A3  E grave
164  10/04  244  A4  E circumflex
165  10/05  245  A5  E diaeresis
166  10/06  246  A6  I circumflex
167  10/07  247  A7  I diaeresis
168  10/08  250  A8  Acute accent
169  10/09  251  A9  Grave accent
170  10/10  252  AA  Circumflex accent
171  10/11  253  AB  Diaeresis
172  10/12  254  AC  Tilde accent
173  10/13  255  AD  U grave
174  10/14  256  AE  U circumflex
175  10/15  257  AF  Lira symbol
176  11/00  260  B0  Top bar (macron)
177  11/01  261  B1  Y acute
178  11/02  262  B2  y acute
179  11/03  263  B3  Degree Sign
180  11/04  264  B4  C cedilla
181  11/05  265  B5  c cedilla
182  11/06  266  B6  N tilde
183  11/07  267  B7  n tilde
184  11/08  270  B8  Inverted exclamation mark
185  11/09  271  B9  Inverted question mark
186  11/10  272  BA  Currency symbol
187  11/11  273  BB  Pound sterling symbol
188  11/12  274  BC  Yen symbol
189  11/13  275  BD  Paragraph
190  11/14  276  BE  Florin (Guilder) symbol
191  11/15  277  BF  Cent symbol
192  12/00  300  C0  a circumflex
193  12/01  301  C1  e circumflex
194  12/02  302  C2  o circumflex
195  12/03  303  C3  u circumflex
196  12/04  304  C4  a acute
197  12/05  305  C5  e acute
198  12/06  306  C6  o acute
199  12/07  307  C7  u acute
200  12/08  310  C8  a grave
201  12/09  311  C9  e grave
202  12/10  312  CA  o grave
203  12/11  313  CB  u grave
204  12/12  314  CC  a diaeresis
205  12/13  315  CD  e diaeresis
206  12/14  316  CE  o diaeresis
207  12/15  317  CF  u diaeresis
208  13/00  320  D0  A ring
209  13/01  321  D1  i circumflex
210  13/02  322  D2  O with stroke
211  13/03  323  D3  AE digraph
212  13/04  324  D4  a ring
213  13/05  325  D5  i acute
214  13/06  326  D6  o with stroke
215  13/07  327  D7  ae digraph
216  13/08  330  D8  A diaeresis
217  13/09  331  D9  i grave
218  13/10  332  DA  O diaeresis
219  13/11  333  DB  U diaeresis
220  13/12  334  DC  E acute
221  13/13  335  DD  i diaeresis
222  13/14  336  DE  German sharp s
223  13/15  337  DF  O circumflex
224  14/00  340  E0  A acute
225  14/01  341  E1  A tilde
226  14/02  342  E2  a tilde
227  14/03  343  E3  Icelandic Eth
228  14/04  344  E4  Icelandic eth
229  14/05  345  E5  I acute
230  14/06  346  E6  I grave
231  14/07  347  E7  O acute
232  14/08  350  E8  O grave
233  14/09  351  E9  O tilde
234  14/10  352  EA  o tilde
235  14/11  353  EB  S caron
236  14/12  354  EC  s caron
237  14/13  355  ED  U acute
238  14/14  356  EE  Y diaeresis
239  14/15  357  EF  y diaeresis
240  15/00  360  F0  Icelandic Thorn
241  15/01  361  F1  Icelandic thorn
242  15/02  362  F2  Middle dot
243  15/03  363  F3  Greek mu
244  15/04  364  F4  Pilcrow sign
245  15/05  365  F5  Fraction 3/4
246  15/06  366  F6  Long dash, horizontal bar
247  15/07  367  F7  Fraction 1/4
248  15/08  370  F8  Fraction 1/2
249  15/09  371  F9  Feminine ordinal
250  15/10  372  FA  Masculine ordinal
251  15/11  373  FB  Left guillemot
252  15/12  374  FC  Solid box
253  15/13  375  FD  Right guillemot
254  15/14  376  FE  Plus or minus sign
255  15/15  377  FF  (Undefined)

    III.1.2. Greek Character Sets

    III.1.2.1. The ISO 8859-7 Latin / Greek Alphabet = ELOT 928

dec col/row oct hex  description
160  10/00  240  A0  No-break space
161  10/01  241  A1  Left single quotation mark
162  10/02  242  A2  right single quotation mark
163  10/03  243  A3  Pound sign
164  10/04  244  A4  (UNUSED)
165  10/05  245  A5  (UNUSED)
166  10/06  246  A6  Broken bar
167  10/07  247  A7  Paragraph sign
168  10/08  250  A8  Diaeresis (Dialytika)
169  10/09  251  A9  Copyright sign
170  10/10  252  AA  (UNUSED)
171  10/11  253  AB  Left angle quotation
172  10/12  254  AC  Not sign
173  10/13  255  AD  Soft hyphen
174  10/14  256  AE  (UNUSED)
175  10/15  257  AF  Horizontal bar (Parenthetiki pavla)
176  11/00  260  B0  Degree sign
177  11/01  261  B1  Plus-minus sign
178  11/02  262  B2  Superscript two
179  11/03  263  B3  Superscript three
180  11/04  264  B4  Accent (tonos)
181  11/05  265  B5  Diaeresis and accent (Dialytika and Tonos)
182  11/06  266  B6  Alpha with accent
183  11/07  267  B7  Middle dot (Ano Teleia)
184  11/08  270  B8  Epsilon with accent
185  11/09  271  B9  Eta with accent
186  11/10  272  BA  Iota with accent
187  11/11  273  BB  Right angle quotation
188  11/12  274  BC  Omicron with accent
189  11/13  275  BD  One half
190  11/14  276  BE  Upsilon with accent
191  11/15  277  BF  Omega with accent
192  12/00  300  C0  iota with diaeresis and accent
193  12/01  301  C1  Alpha
194  12/02  302  C2  Beta
195  12/03  303  C3  Gamma
196  12/04  304  C4  Delta
197  12/05  305  C5  Epsilon
198  12/06  306  C6  Zeta
199  12/07  307  C7  Eta
200  12/08  310  C8  Theta
201  12/09  311  C9  Iota
202  12/10  312  CA  Kappa
203  12/11  313  CB  Lamda
204  12/12  314  CC  Mu
205  12/13  315  CD  Nu
206  12/14  316  CE  Ksi
207  12/15  317  CF  Omicron
208  13/00  320  D0  Pi
209  13/01  321  D1  Rho
210  13/02  322  D2  (UNUSED)
211  13/03  323  D3  Sigma
212  13/04  324  D4  Tau
213  13/05  325  D5  Upsilon
214  13/06  326  D6  Phi
215  13/07  327  D7  Khi
216  13/08  330  D8  Psi
217  13/09  331  D9  Omega
218  13/10  332  DA  Iota with diaeresis
219  13/11  333  DB  Upsilon with diaeresis
220  13/12  334  DC  alpha with accent
221  13/13  335  DD  epsilon with accent
222  13/14  336  DE  eta with accent
223  13/15  337  DF  iota with accent
224  14/00  340  E0  upsilon with diaeresis and accent
225  14/01  341  E1  alpha
226  14/02  342  E2  beta
227  14/03  343  E3  gamma
228  14/04  344  E4  delta
229  14/05  345  E5  epsilon
230  14/06  346  E6  zeta
231  14/07  347  E7  eta
232  14/08  350  E8  theta
233  14/09  351  E9  iota
234  14/10  352  EA  kappa
235  14/11  353  EB  lamda
236  14/12  354  EC  mu
237  14/13  355  ED  nu
238  14/14  356  EE  ksi
239  14/15  357  EF  omicron
240  15/00  360  F0  pi
241  15/01  361  F1  rho
242  15/02  362  F2  terminal sigma
243  15/03  363  F3  sigma
244  15/04  364  F4  tau
245  15/05  365  F5  upsilon
246  15/06  366  F6  phi
247  15/07  367  F7  khi
248  15/08  370  F8  psi
249  15/09  371  F9  omega
250  15/10  372  FA  iota with diaeresis
251  15/11  373  FB  upsilon with diaeresis
252  15/12  374  FC  omicron with diaeresis
253  15/13  375  FD  upsilon with accent
254  15/14  376  FE  omega with accent
255  15/15  377  FF  (UNUSED)

    III.1.2.2. The ELOT 927 Character Set

dec col/row oct hex  description
 32  02/00   40  20  SPACE
 33  02/01   41  21  EXCLAMATION MARK
 34  02/02   42  22  QUOTATION MARK
 35  02/03   43  23  NUMBER SIGN
 36  02/04   44  24  DOLLAR SIGN
 37  02/05   45  25  PERCENT SIGN
 38  02/06   46  26  AMPERSAND
 39  02/07   47  27  APOSTROPHE
 40  02/08   50  28  LEFT PARENTHESIS
 41  02/09   51  29  RIGHT PARENTHESIS
 42  02/10   52  2A  ASTERISK
 43  02/11   53  2B  PLUS SIGN
 44  02/12   54  2C  COMMA
 45  02/13   55  2D  HYPHEN, MINUS SIGN
 46  02/14   56  2E  PERIOD, FULL STOP
 47  02/15   57  2F  SOLIDUS, SLASH
 48  03/00   60  30  DIGIT ZERO
 49  03/01   61  31  DIGIT ONE
 50  03/02   62  32  DIGIT TWO
 51  03/03   63  33  DIGIT THREE
 52  03/04   64  34  DIGIT FOUR
 53  03/05   65  35  DIGIT FIVE
 54  03/06   66  36  DIGIT SIX
 55  03/07   67  37  DIGIT SEVEN
 56  03/08   70  38  DIGIT EIGHT
 57  03/09   71  39  DIGIT NINE
 58  03/10   72  3A  COLON
 59  03/11   73  3B  SEMICOLON
 60  03/12   74  3C  LESS-THAN SIGN, LEFT ANGLE BRACKET
 61  03/13   75  3D  EQUALS SIGN
 62  03/14   76  3E  GREATER-THAN SIGN, RIGHT ANGLE BRACKET
 63  03/15   77  3F  QUESTION MARK
 64  04/00  100  40  COMMERCIAL AT SIGN
 65  04/01  101  41  CAPITAL LETTER A
 66  04/02  102  42  CAPITAL LETTER B
 67  04/03  103  43  CAPITAL LETTER C
 68  04/04  104  44  CAPITAL LETTER D
 69  04/05  105  45  CAPITAL LETTER E
 70  04/06  106  46  CAPITAL LETTER F
 71  04/07  107  47  CAPITAL LETTER G
 72  04/08  110  48  CAPITAL LETTER H
 73  04/09  111  49  CAPITAL LETTER I
 74  04/10  112  4A  CAPITAL LETTER J
 75  04/11  113  4B  CAPITAL LETTER K
 76  04/12  114  4C  CAPITAL LETTER L
 77  04/13  115  4D  CAPITAL LETTER M
 78  04/14  116  4E  CAPITAL LETTER N
 79  04/15  117  4F  CAPITAL LETTER O
 80  05/00  120  50  CAPITAL LETTER P
 81  05/01  121  51  CAPITAL LETTER Q
 82  05/02  122  52  CAPITAL LETTER R
 83  05/03  123  53  CAPITAL LETTER S
 84  05/04  124  54  CAPITAL LETTER T
 85  05/05  125  55  CAPITAL LETTER U
 86  05/06  126  56  CAPITAL LETTER V
 87  05/07  127  57  CAPITAL LETTER W
 88  05/08  130  58  CAPITAL LETTER X
 89  05/09  131  59  CAPITAL LETTER Y
 90  05/10  132  5A  CAPITAL LETTER Z
 91  05/11  133  5B  LEFT SQUARE BRACKET
 92  05/12  134  5C  REVERSE SOLIDUS, BACKSLASH
 93  05/13  135  5D  RIGHT SQUARE BRACKET
 94  05/14  136  5E  CIRCUMFLEX ACCENT
 95  05/15  137  5F  UNDERSCORE
 96  06/00  140  60  ACCENT GRAVE
 97  06/01  141  61  GREEK LETTER ALPHA
 98  06/02  142  62  GREEK LETTER BETA
 99  06/03  143  63  GREEK LETTER GAMMA
100  06/04  144  64  GREEK LETTER DELTA
101  06/05  145  65  GREEK LETTER EPSILON
102  06/06  146  66  GREEK LETTER ZETA
103  06/07  147  67  GREEK LETTER ETA
104  06/08  150  68  GREEK LETTER THETA
105  06/09  151  69  GREEK LETTER IOTA
106  06/10  152  6A  GREEK LETTER KAPPA
107  06/11  153  6B  GREEK LETTER LAMDA
108  06/12  154  6C  GREEK LETTER MU
109  06/13  155  6D  GREEK LETTER NU
110  06/14  156  6E  GREEK LETTER KSI
111  06/15  157  6F  GREEK LETTER OMICRON
112  07/00  160  70  GREEK LETTER PI
113  07/01  161  71  GREEK LETTER RHO
114  07/02  162  72  GREEK LETTER SIGMA
115  07/03  163  73  GREEK LETTER TAU
116  07/04  164  74  GREEK LETTER UPSILON
117  07/05  165  75  GREEK LETTER FI
118  07/06  166  76  GREEK LETTER XI
119  07/07  167  77  GREEK LETTER PSI
120  07/08  170  78  GREEK LETTER OMEGA
121  07/09  171  79  SPACE
122  07/10  172  7A  SPACE
123  07/11  173  7B  LEFT CURLY BRACKET, LEFT BRACE
124  07/12  174  7C  VERTICAL LINE, VERTICAL BAR
125  07/13  175  7D  RIGHT CURLY BRACKET, RIGHT BRACE
126  07/14  176  7E  TILDE
127  07/15  177  7F  RUBOUT, DELETE

    III.1.2.3. PC Code Page 869

   (to be filled in...)

    III.2. Updated Country Codes

   Date: Mon, 7 Apr 1997 23:23:49 EDT
   From: Dave Leibold <dleibold@else.net>
   Newsgroups: comp.dcom.telecom
   Subject: Ex-USSR Country Codes Profile
   Organization: TELECOM Digest

   Ex-USSR Country Codes Profile
   4 April 1997

   Below is a summary of the country codes that have formed in the wake of
   the USSR dissolution, along with some updated findings and reports.
   Additional or corrected information on any of these nations would be
   welcome (c/o dleibold@else.net).
     * Kyrgyz Republic country code 996 will take effect, at least in
       Canada, effective 1 May 1997, according to CRTC Telecom Order
       97-464, based on Stentor Tariff Notice 433. There is no indication
       whether there will be a permissive dialing period involved or for
       how long such a permissive operation would remain.
     * Country code 992 was reported as a recent assignment for
       Tajikistan, which will be moving from country code 7 at some
       unknown time.
     * Uzbekistan has its own country code assignment, but I have no
       information if this is in service yet or what implementation dates
       have been set.
     * Kazakstan does not have a known separate country code assignment at
       present. It remains in country code 7 for the time being.
     * Russia seems destined to keep country code 7.
     * Recent news reports speak of some agreements forming between Russia
       and Belarus. While there is no outright reunification yet, there is
       expected to be much closer ties between the two nations. Whether
       this will lead to a reunification of telephone codes remains to be
       seen.

   In the table, "Effective" means the date at which the country code
   began service (which could vary according to the nation). "Mandatory"
   means the date at which the country code 7 is invalid for calls to that
   nation. There are a number of question marks since exact dates have not
   been collected in all cases.

CC  Nation            Effective     Mandatory    Notes

370 Lithuania         1993?         ???          Announced Jan 1993
371 Latvia            1993?         ???
372 Estonia           1 Feb 1993?   March 1993?
373 Moldova           1993?         ???          Announced Jan 1993
374 Armenia           1 May 1995    1 July 1995  Announced Jan 1995 (ITU)
375 Belarus           16 Apr 1995   1997?
380 Ukraine           16 Apr 1995   Oct 1995?
7   Kazakstan         (no known changes)
7   Russia            (presumably not changing)
992 Tajikistan        ???           ???          Announced 1996-7?
993 Turkmenistan      3 Jan 1997    3 Apr 1997   Canada as of 29 Nov 1996
994 Azerbaijan        Sept 1994?    ???          Announced 1992
995 Georgia           1994?         ???          ref: Telecom Digest Oct 1994
996 Kyrgyz Republic   1 May 1997    ???          ref: Stentor Canada/CRTC
998 Uzbekistan        ???           ???          Announced 1996? (ITU)

   Details courtesy Toby Nixon, ITU, Stentor (Canada), CRTC (Canada),
   TELECOM Digest (including information collected for the country code
   listings).

IV. ERRATA & CORRIGENDA

   The following errors inUsing C-Kermit, Second Edition, first
   printing, have been noted.

   First, some missing acknowledgements for C-Kermit 6.0: JE Jones of
   Microware for help with OS-9, Nigel Roles for his help with Plan 9,
   Lucas Hart for help with VMS and Digital UNIX, Igor Kovalenko for his
   help with QNX. And later, to Susan Kleinmann for her help with Debian
   Linux packaging; Patrick Volkerding for his help with Slackware Linux
   packaging; Jim Knoble for his help with Red Hat Linux packaging; and to
   dozens of others for sending individual C-Kermit binaries for varied
   and diverse platforms.

   Thanks to James Spath for both binaries and reporting many of the typos
   noted below. Also to Dat Thuc Nguyen for spotting several typos.

PAGE    REMARKS
COVER   "COS" is a misprint.  There is no COS.  Pretend it says "SCO" or "VOS".
        (This is fixed in the second printing.)
 xxi    Second line: Fred Smith's affiliation should be Computrition.
 83     Change "commands other" to "commands as other" (1st paragraph)
 87     Change "The the" to "The" (2nd paragraph)
 92     "set modem-type user-defined supra" should be "set modem type ..."
 95     Change "VI" to "vi" (1st paragraph)
 96     Change "it it" to "it is" (1st paragraph)
 97     Change "advantage a literal" to "advantage of a literal" (2nd
        paragraph)
102     The call-waiting example would be better as SET DIAL PREFIX *70W
        (rather than "*70,") because the former will not cause an incorrect
        call to be placed with pulse dialing.
123     Third paragraph from bottom: "..otherwise if a your local username.."
        should be "..otherwise your local username..".
160     Delete the "it" between "and" and "to" (2nd paragraph)
185     In "When TRANSFER DISPLAY is OFF, C-Kermit skips the display...",
        "OFF" should be "NONE".
187     The last paragraph says the "A command" is ignored, should be "S".
194     Change "it known" to "it is known" (4th paragraph).
235     In C-Kermit 7.0, the syntax of the GET command changed.  MGET now
        must be used to get a list of files and there is no more multiline
        GET command.
268     Last paragraph: "effect" should be "affect".
275     In the SET PROTOCOL KERMIT description, the following sentence is
        incorrect and should be removed: 'If you omit the commands, the
        default ones are restored: "kermit -ir" and "kermit -r" respectively".
        The correct information is given at the bottom of page 281.
279     9th line.  The decimal value of ST is 156, not 155.
295     In the stepping stones, skip ahead to Chapter 17 on p. 327.
298     Table 16-2, Portuguese entry.  Column 4/00 should show section sign,
        not acute accent.
316     Other languages written in the Hebrew alphabet include Karaim (a Turkic
        language spoken in Lithuania and Poland), Judeo-Kurdish, and Judeo-
        Georgian.
332     UNDEFINE definition, change "This just" to "This is just".
344     It might be necessary to set the modem's pulse generation rate when
        sending numeric pages; most Hayes compatible modems use the S11
        register for this.
350     Delete "is" from between "It" and "ceases" (4th paragraph)
351     Top - both occurrences of "print \%a" should be "echo \%a".
364     \v(input) and \v(query) out of alphabetical order.
378     In the MYSEND macro, "if not \m(rc) goto bad" should be:
        "if \m(rc) goto bad" (remove the "not").
382-383 It should be stated that the loop control variable must be of the \%a
        type, or else an array element; macro names can not be used for this.
383     In line 3, "\%f[\%i]" should be "\&f[\%i]".
383     In the sort example, it should be stated that the array is 1-based.
387     Change "You can list" to "You can get a list" (5th paragraph)
393     \Fverify() description.  The 3rd sentence could be stated more clearly
        as "If all characters in string2 are also in string1, 0 is returned."
398     Copying \ffiles() results to an array before is not required as of
        C-Kermit 7.0 (see Section 7.3).
403     In "(\%a + 3) * (\%b  5)", a minus sign is missing between b and 5.
407     C-Kermit 7.0 no longer supports multiline GET.  Change
        "get, \%1, \%2" to "get {\%1} {\%2}" or "get /as:{\%2} {\%1}".
409     READ example while loop should be:
        while success { echo \m(line), read line }
409     "WRITE file" should be "WRITE keyword" (you can't put a filename there)
        (The same applies to WRITE-LINE / WRITELN).
414     \Funhexify() missing from Table 18-3.
425     MINPUT definition, change 2nd "text2" to "text3".
436     Several lines are missing from the UNIXLOGIN macro listing.
        After the "xif fail" block, insert:

          out \%1\13                    ; Send username, carriage return
          inp 5 Password:               ; Wait 5 sec for this prompt
          if fail end 1 No password prompt
          pause                         ; Wait a sec
          out \%2\13                    ; Send password

440     Change "set terminal byteszie" to "set terminal bytesize".
        Change "input Password:" to "input 10 Password".
448     Franchise script: "access line" should be "access \m(line)".
453     There are two incorrectly coded IF statements in the DELIVER macro
        definition.  Replace both occurrences of "if > \%1 \%3 {" with
        "xif > \%i \%3 {" (replace "if" by "xif" and "\%1" with "\%i").
453     "the the" (last paragraph) should be "the".
454     EOT (last paragraph) is End of Transmission, not End of Text.
457     _DEFINE definition: "name constructed" should be "name is constructed".
457     "macro for and" (last paragraph) should be "macro and".
459     Should explain that \v(user) is a legal abbreviation of \v(userid).
480     Figure II-2 is backwards; the least-significant bit is transmitted
        first, then up to the highest, and the parity bit last.
534     The VMS Appendix section on Odd Record Lengths no longer applies;
        C-Kermit 7.0 handles odd record lengths as well as even ones.
559     Table VIII-3, Portuguese entry.  Column 4/00 should show section sign,
        not acute accent.
560-563 HP-Roman8 missing from Table VII-4; there wasn't room to squeeze it in.
        It is listed in section II(6).
565     "d stroke" in Table VII-5 has the wrong appearance; the stem should
        be upright.  The letter shown in the table is actually a lowercase
        Icelandic eth, which has a curved stem.
601-604 BeBox, BeOS, Plan 9, and probably others not listed in trademarks.
604     The words "SCRIBE TEXT FORMATTER" appear at the end of the last
        sentence of the first paragraph of the Colophon.  They should have
        been in the Index.
Index:  Missing entries: SET { SEND, RECEIVE } PATHNAMES, Call waiting, ...
        \F()            Page 605, add also 413-414
        \Fbreak         389
        \Fcapitalize    390
        \Fchecksum      414
        \Fcrc16         414
        \Fexecute       414
        \Fhexify        390
        \Fltrim         391
        \Frepeat        392
        \Fspawn         392
        \Ftod2secs      399
        \v() built_in   Page 606, add also 361-364
        \v(_line)       354, 361
        \v(apcactive)   361
        \v(charset)     362
        \v(cpu)         362
        \v(crc16)       357, 362
        \v(d$xxx)       add page 362
        \v(dialnumber)  362
        \v(dialresult)  362
        \v(errno)       362
        \v(errstring)   362
        \v(exedir)      362
        \v(inidir)      363
        \v(ipaddress)   363
        \v(keyboard)    363
        \v(macro)       363
        \v(minput)      363
        \v(m_xxx)       94, 363
        \v(password)    364
        \v(query)       364
        \v(prompt)      364
        \v(speed)       356, 364
        \v(startup)     364
        \v(status)      364
        \v(sysid)       364
        \v(system)      364
        \v(fsize)       at lower half page 606 should read \v(tfsize)
        \v(xversion)    364
        BEEP Command    40
        SET FLOW        62, 212

   Figure II-5 on page 493. The pin assignments of the Mini Din-8
   connector are not described anywhere. As noted in the text, these tend
   to vary from vendor to vendor. One common arrangement is:

  1. HSKout (Handshake out -- definition depends on software)
  2. HSKin  (Handshake in or external clock)
  3. TxD-
  4. Not used
  5. RxD-
  6. TxD+
  7. Not used
  8. RxD+

   Note the "balanced pairs" for Receive Data (RxD) and Transmit Data
   (TxD), and the utter lack of modem signals. These connectors follow the
   RS-423 standard, rather than RS-232. In some arrangements, Pin 1 is
   used for DTR and Pin 2 for CD; in others Pin 1 is RTS and Pin 2 is CTS.

   Please send reports of other errors to the authors, as well as
   suggestions for improvements, additional index entries, and any other
   comments:

  kermit@columbia.edu

APPENDIX V. ADDITIONAL COPYRIGHT NOTICES

   The following copyrights cover some of the source code used in the
   development of C-Kermit, Kermit 95, or Kermit 95 support libraries.

/*****************************************************************************/
/*                                                                           */
/*              Copyright (c) 1995 by Oy Online Solutions Ltd.               */
/*                                                                           */
/*   Distribution of this source code is strictly forbidden. Use of this    */
/*   source code is granted to the University of Columbia C-Kermit project   */
/*   to be distributed in binary format only. Please familiarize yourself    */
/*   with the accompanying LICENSE.P file.                                   */
/*                                                                           */
/*****************************************************************************/

   used for Xmodem, Ymodem, and Zmodem protocol in Kermit 95 (p95.dll,
   p2.dll)

   Copyright (c) 1997 Stanford University

   The use of this software for revenue-generating purposes may require a
   license from the owners of the underlying intellectual property.
   Specifically, the SRP-3 protocol may not be used for revenue-generating
   purposes without a license.

   Within that constraint, permission to use, copy, modify, and distribute
   this software and its documentation for any purpose is hereby granted
   without fee, provided that the above copyright notices and this
   permission notice appear in all copies of the software and related
   documentation.

   THE SOFTWARE IS PROVIDED "AS-IS" AND WITHOUT WARRANTY OF ANY KIND,
   EXPRESS, IMPLIED OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY
   WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

   IN NO EVENT SHALL STANFORD BE LIABLE FOR ANY SPECIAL, INCIDENTAL,
   INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY KIND, OR ANY DAMAGES
   WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER OR NOT
   ADVISED OF THE POSSIBILITY OF DAMAGE, AND ON ANY THEORY OF LIABILITY,
   ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
   SOFTWARE.

   Used for Secure Remote Password (TM) protocol (SRP) in C-Kermit, Kermit
   95 (k95.exe, k2.exe, k95crypt.dll, k2crypt.dll)

   Copyright 1990 by the Massachusetts Institute of Technology. All Rights
   Reserved.

   Export of this software from the United States of America may require a
   specific license from the United States Government. It is the
   responsibility of any person or organization contemplating export to
   obtain such a license before exporting.

   WITHIN THAT CONSTRAINT, permission to use, copy, modify, and distribute
   this software and its documentation for any purpose and without fee is
   hereby granted, provided that the above copyright notice appear in all
   copies and that both that copyright notice and this permission notice
   appear in supporting documentation, and that the name of M.I.T. not be
   used in advertising or publicity pertaining to distribution of the
   software without specific, written prior permission. M.I.T. makes no
   representations about the suitability of this software for any purpose.
   It is provided "as is" without express or implied warranty.

   Used for Telnet Authentication Option, Telnet Encryption Option, and
   Kerberos (TM) authentication in C-Kermit, Kermit 95 (k95.exe, k2.exe,
   k95crypt.dll, k2crypt.dll)

   Copyright (c) 1991, 1993 The Regents of the University of California.
   All rights reserved.

   Redistribution and use in source and binary forms, with or without
   modification, are permitted provided that the following conditions are
   met:

    1. Redistributions of source code must retain the above copyright
       notice, this list of conditions and the following disclaimer.
    2. Redistributions in binary form must reproduce the above copyright
       notice, this list of conditions and the following disclaimer in the
       documentation and/or other materials provided with the
       distribution.
    3. All advertising materials mentioning features or use of this
       software must display the following acknowledgement:

     This product includes software developed by the University of
     California, Berkeley and its contributors.

    4. Neither the name of the University nor the names of its
       contributors may be used to endorse or promote products derived
       from this software without specific prior written permission.

   THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
   PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS
   BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
   CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
   SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
   BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
   WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
   OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
   ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

   Used for Telnet Authentication Option, Telnet Encryption Option, and
   Kerberos (TM) authentication in C-Kermit, Kermit 95 (k95.exe, k2.exe,
   k95crypt.dll, k2crypt.dll)

   Copyright (C) 1995-1997 Eric Young (eay@cryptsoft.com) All rights
   reserved.

   This package is an DES implementation written by Eric Young
   (eay@cryptsoft.com). The implementation was written so as to conform
   with MIT's libdes.

   This library is free for commercial and non-commercial use as long as
   the following conditions are aheared to. The following conditions apply
   to all code found in this distribution.

   Copyright remains Eric Young's, and as such any Copyright notices in
   the code are not to be removed. If this package is used in a product,
   Eric Young should be given attribution as the author of that the SSL
   library. This can be in the form of a textual message at program
   startup or in documentation (online or textual) provided with the
   package.

   Redistribution and use in source and binary forms, with or without
   modification, are permitted provided that the following conditions are
   met:

    1. Redistributions of source code must retain the copyright notice,
       this list of conditions and the following disclaimer.
    2. Redistributions in binary form must reproduce the above copyright
       notice, this list of conditions and the following disclaimer in the
       documentation and/or other materials provided with the
       distribution.
    3. All advertising materials mentioning features or use of this
       software must display the following acknowledgement: This product
       includes software developed by Eric Young (eay@cryptsoft.com)

   THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND ANY EXPRESS OR
   IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
   WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
   DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
   ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
   STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
   IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
   POSSIBILITY OF SUCH DAMAGE.

   The license and distribution terms for any publically available version
   or derivative of this code cannot be changed. i.e. this code cannot
   simply be copied and put under another distribution license [including
   the GNU Public License.]

   The reason behind this being stated in this direct manner is past
   experience in code simply being copied and the attribution removed from
   it and then being distributed as part of other packages. This
   implementation was a non-trivial and unpaid effort.

   Used DES encryption in Kermit 95 (k95crypt.dll, k2crypt.dll)
     __________________________________________________________________

 * This is version 1.1 of CryptoLib
 *
 * The authors of this software are Jack Lacy, Don Mitchell and Matt Blaze
 *              Copyright (c) 1991, 1992, 1993, 1994, 1995 by AT&T.
 * Permission to use, copy, and modify this software without fee
 * is hereby granted, provided that this entire notice is included in
 * all copies of any software which is or includes a copy or
 * modification of this software and in all copies of the supporting
 * documentation for such software.
 *
 * NOTE:
 * Some of the algorithms in cryptolib may be covered by patents.
 * It is the responsibility of the user to ensure that any required
 * licenses are obtained.
 *
 *
 * SOME PARTS OF CRYPTOLIB MAY BE RESTRICTED UNDER UNITED STATES EXPORT
 * REGULATIONS.
 *
 *
 * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
 * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR AT&T MAKE ANY
 * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
 * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.

   Used for Big Number library in Kermit 95 (k95crypt.dll, k2crypt.dll).

     __________________________________________________________________

   CKERMIT70.TXT
   The Kermit Project
   Columbia University
   8 Feb 2000
   Last update: 8 Aug 2011