stdlib/Sockets: getaddrinfo("☃.net") non-ASCII hostname (RFC 3492)

[samoconnor]

julia/stdlib/Sockets/src/addrinfo.jl

Lines 51 to 52 in 569981a

	function getalladdrinfo(host::String)
	isascii(host) || error("non-ASCII hostname: $host")

Should this changed be to isascii(host) || host = Punycoder.encode(host) ?

References:
https://www.ietf.org/rfc/rfc3492.txt
JuliaWeb/URIParser.jl#9
http://☃.net

[vtjnash]

Seems vaguely like something that should be handled at the application-level (URIParser.jl? HTTP.jl? higher?), and not in the stdlib.

[samoconnor]

I'm not familiar with this in practice, just noticed this as a loose end in the URIParser.jl issues.
However, it seems like "Internationalized Domain Names in Applications (IDNA): Protocol" https://tools.ietf.org/html/rfc5891, is a well defined thing. It says DNS lookup of non-ascii names should be done by converting to punycode (an "A-label" in the language of the RFC).

5.6.  DNS Name Resolution

The A-label resulting from the conversion in Section 5.5 or supplied
directly (see Section 5.3) is combined with other labels as needed to
form a fully-qualified domain name that is then looked up in the DNS,
using normal DNS resolver procedures.  The lookup can obviously
either succeed (returning information) or fail.

Doing this in stdlib/Sockets might be a simple way to help all Julia systems to be IDNA compatible.

[inkydragon]

upstream libuv has already supported non-ASCII hostname.

• unix,win: support IDNA 2008 in uv_getaddrinfo() · libuv/libuv@6dd44ca

so, just remove this line and getaddrinfo("☃.net") works fine.

julia/stdlib/Sockets/src/addrinfo.jl

Line 60 in 4f8a7b9

isascii(host) || error("non-ASCII hostname: $host")

[inkydragon]

using Sockets; getaddrinfo("☃.net") works fine on Julia 1.6.3/1.7.0-rc1.

This issue can be closed now.