Skip to content

Podcastindex-org/podping.cloud

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Podping.cloud

Podping.cloud is the hosted front-end to the Podping notification system. It stands in front of the back-end writer(s) to provide a more friendly HTTP based API.


Overview

There are two main components of a podping.cloud node. The first is a web HTTP front-end just called podping that accepts GET requests like so:

GET https://podping.cloud/?url=https://feeds.example.org/podcast/rss

You can also append 2 additional parameters (reason and/or medium):

GET https://podping.cloud/?url=https://feeds.example.org/livestream/rss&reason=live&medium=music

If reason is not present, the default is "update". If medium is not present, the default is "podcast". A full explanation of these options and what they mean
is here.

The next component is one or more back-end writers that connect to the front-end over a ZMQ socket. Currently, the only back-end writer is the hive-writer, a python script that listens on localhost port 9999 for incoming events. When it receives an event, it attempts to write that event as a custom JSON notification message to the Hive blockchain.


Requests

The front-end accepts GET requests and does a few things:

  1. Ensures that the sending publisher has included a valid 'Authorization' header token.
  2. Validates that the token exists in the auth.db sqlite db.
  3. Validates that the format of the given podcast feed url looks sane
  4. Saves the url into the queue.db sqlite database in the queue table.
  5. Returns 200 to the sending publisher.

A separate thread runs in a loop as a queue checker and does the following:

  1. Checks the queue.db database and fetches up to 1000 feeds at a time in FIFO order ("live" reason is prioritized).
  2. Checks the ZEROMQ tcp socket to the hive-writer listener on port 9999.
  3. Construct one or more Ping objects in protocol buffers to send over the socket to the writer(s).
  4. Sends the Ping objects to hive-writer socket for processing and waits for success or error to be returned.
  5. If success is returned from the writer, the url is removed from queue.db.
  6. If an error is returned or an exception is raised, another attempt is made after 180 seconds.

There is a dummy auth token in the auth.db that is ready to use for testing. The token value is:


Blahblah^^12345678

In order to avoid running as a root user, please set the PODPING_RUNAS_USER environment variable to the non-root user you want the front-end executable to run as. Something like this:

PODPING_RUNAS_USER="podping" ./target/release/podping

Back-end Writers

  • Podping-hivewriter: Accepts events from the podping.cloud front-end or from the command line and writes them to the Hive blockchain.

Running a Full Podping.cloud Node

The best way to run a podping.cloud node is with docker compose. There is a [docker] folder for this. Just clone this repo, switch to the docker folder and issue docker compose up. It is expected that the database files will live in a directory called /data. If this directory doesn't exist, you will need to create it.

Initially, the auth.db and queue.db will be blank. You will need to populate the "publishers" table in the auth.db file to have a funcional system. See the example files in the databases directory in this repo for an example of the format for publisher token records.


The Podping Network Idea

Framework Overview 1