Triggered web site rebuilds and similar

[Owen Taylor <otaylor redhat com>]

Right now to rebuild web-sites for gnome.org from version control we
have a setup that looks something like:

 - we send mail to a user on an a different machine
 - in the procmailrc for that user we touch a file
 - in a cron job we notice that the file has changed and perform
   the rebuild

That's pretty horrible. A little while ago I sat down over an evening
or two and created a simple daemon/client pair called 'fishpoll'. It
basically works like this:

 A) On window.gnome.org you install fishpolld and a script

     /etc/fishpoll.d/rebuild_gnome_web

 B) In your commit hook, you do:

     fishpollc window.gnome.org rebuild_gnome_web

(You'd probably actually do it over the back-channel network, not the
public network, simplifying above.) That talks to fishpolld on the
other machine and tells it to run the script at the next opportunity.
It will never run the same handler twice at the same time so its
self-throttling.

There's more details at:

  http://git.fishsoup.net/cgit/fishpoll/tree/README
  git clone git://git.fishsoup.net/fishpoll

I'm thinking of deploying this on gnome.org, at least for a few things,
but have some feeling I'm probably reinventing the wheel badly. It's an
obvious problem, there must be a better standard answer... or is there?

Suggestions appreciated.

- Owen