A simple tool that parses content feeds and sends out appropriate push notifications (WebSub, webmention, etc.) when they change.

See http://publ.beesbuzz.biz/blog/113-Some-thoughts-on-WebMention for the motivation.

• Supports any feed supported by feedparser and mf2py (RSS, Atom, HTML pages containing h-entry, etc.)
• Will send WebSub notifications for feeds which declare a WebSub hub
• Will send WebMention notifications for entries discovered on those feeds or specified directly
• Can perform autodiscovery of additional feeds on entry pages
• Can do a full backfill on Atom feeds configured with RFC 5005
• When configured to use a cache directory, can detect entry deletions and updates to implement the webmention update and delete protocols (as well as saving some time and bandwidth)

If you want to support WebSub, have your feed implement the WebSub protocol. The short version is that you should have a <link rel="hub" href="http://path/to/hub" /> in your feed's top-level element.

There are a number of WebSub hubs available; I use Superfeedr.

For WebMentions, configure your site templates with the various microformats; by default, Pushl will use the following tags as the top-level entry container, in descending order of priority:

• Anything with a class of h-entry
• An <article> tag
• Anything with a class of entry

For more information on how to configure your site templates, see the microformats h-entry specification.

If you're using an mf2 feed (i.e. an HTML-formatted page with h-entry declarations), only entries with a u-url property will be used for sending webmentions; further, Pushl will retrieve the page from that URL to ensure it has the full content. (This is to work around certain setups where the h-feed only shows summary text.)

Also, there is technically no requirement for an HTML page to declare an h-feed; all entities marked up with h-entry will be consumed.

You can install it using pip with e.g.:

pip3 install pushl

However, I recommend installing it in a virtual environment with e.g.:

virtualenv3 $HOME/pushl
$HOME/pushl/bin/pip3 install pushl

and then putting a symlink to $HOME/pushl/bin/pushl to a directory in your $PATH, e.g.

ln -s $HOME/pushl/bin/pushl $HOME/bin/pushl

pushl -c $HOME/var/pushl-cache http://example.com/feed.xml

While you can run it without the -c argument, its use is highly recommended so that subsequent runs are both less spammy and so that it can detect changes and deletions.

If you just want to send webmentions from an entry page without processing an entire feed, the -e/--entry flag indicates that the following URLs are pages or entries, rather than feeds; e.g.

pushl -e http://example.com/some/page

will simply send the webmentions for that page.

The -r/--recurse flag will discover any additional feeds that are declared on entries and process them as well. This is useful if you have per-category feeds that you would also like to send WebSub notifications on. For example, my site has per-category feeds which are discoverable from individual entries, so pushl -r http://beesbuzz.biz/feed will send WebSub notifications for all of the categories which have recent changes.

Note that -r and -e in conjunction will also cause the feed declared on the entry page to be processed further. While it is tempting to use this in a feed autodiscovery context e.g.

pushl -re http://example.com/blog/

this will also send webmentions from the blog page itself which is probably not what you want to have happen.

If your feed implements RFC 5005, the -a flag will scan past entries for WebMention as well. It is recommended to only use this flag when doing an initial backfill, as it can end up taking a long time on larger sites (and possibly make endpoint operators very grumpy at you). To send updates of much older entries it's better to just use -e to do it on a case-by-case basis.

If you have a website which has multiple URLs that can access it (for example, http+https, or multiple domain names), you generally only want WebMentions to be sent from the canonical URL. The best solution is to use <link rel="canonical"> to declare which one is the real one, and Pushl will use that in sending the mentions; so, for example:

pushl -r https://example.com/feed http://example.com/feed http://alt-domain.example.com/feed

As long as both http://example.com and http://alt-domain.example.com declare the https://example.com version as canonical, only the webmentions from https://example.com will be sent.

If, for some reason, you can't use rel="canonical" you can use the -s/--websub-only flag on Pushl to have it only send WebSub notifications for that feed; for example:

pushl -r https://example.com/feed -s https://other.example.com/feed

will send both Webmention and WebSub for https://example.com but only WebSub for https://other.example.com.

pushl can be run from a cron job, although it's a good idea to use flock -n to prevent multiple instances from stomping on each other. An example cron job for updating a site might look like:

*/5 * * * * flock -n $HOME/.pushl-lock pushl -rc $HOME/.pushl-cache http://example.com/feed

In my setup, I have pushl installed in my website's pipenv:

cd $HOME/beesbuzz.biz
pipenv install pushl

and created this script as $HOME/beesbuzz.biz/pushl.sh:

#!/bin/bash

cd $(dirname "$0")
LOG=logs/pushl-$(date +%Y%m%d.log)

# redirect log output
if [ "$1" == "quiet" ] ; then
    exec >> $LOG 2>&1
else
    exec 2>&1 | tee -a $LOG
fi

# add timestamp
date

# run pushl
flock -n $HOME/var/pushl/run.lock $HOME/.local/bin/pipenv run pushl -rvvkc $HOME/var/pushl \
    https://beesbuzz.biz/feed\?push=1 \
    http://publ.beesbuzz.biz/feed\?push=1 \
    https://tumblr.beesbuzz.biz/rss \
    https://novembeat.com/feed\?push=1 \
    http://beesbuzz.biz/feed\?push=1 \
    -s http://beesbuzz.biz/feed-summary https://beesbuzz.biz/feed-summary

# while we're at it, clean out the log and pushl cache directory
find logs $HOME/var/pushl -type f -mtime +30 -print -delete

Then I have a cron job:

*/15 * * * * $HOME/beesbuzz.biz/pushl.sh quiet

which runs it every 15 minutes.

I also have a git deployment hook for my website, and its final step (after restarting gunicorn) is to run pushl.sh, in case a maximum latency of 15 minutes just isn't fast enough.