There's this relatively new protocol, PubSubHubbub, in which a server publishing an RSS feed pings a server whenever it publishes an update to the feed's XML file. Feed aggregators, such as Google Reader and others, are then notified immediately when the updated feed is available, and can thus refresh it immediately, rather than wait for some timed cronjob to do so. With respect to Planet Fedora, there are 2 things we _could_ do to make it more timely. Currently, planet.fp.o gets updated every 20 minutes by cronjob, rescanning all its feeds. 1) If those feeds were themselves publishing their PubSubHubbub address, we could: not rescan such feeds every 20 minutes, but only when notified that they have new content (plus say daily to be sure we don't miss something). WordPress and others have a plugin to ping a hub, so that's easy for our users to do, and they may already be doing so. 2) Every time we finish publishing an updated aggregated RSS feed, we add a 'ping' to the public PubSubHubbub servers. Doing so, aggregators could then immediately pull our updated feed. >From a planet.fp.o publisher perspective, it's really simple. 1) Include a couple bits in the RSS feed itself: an atom namespace reference, and in the <channel>, an <atom:link> that references the hub server. In this way, each aggregator can look at that atom:link tag and configure itself to subscribe to the pings when those feeds are updated. This should be trivial to patch into Venus, our current planet software. 2) Ping a hub. Here's a python module to do it: http://pypi.python.org/pypi/PubSubHubbub_Publisher/1.0 (other languages available too: http://code.google.com/p/pubsubhubbub/wiki/PublisherClients ). This could be done in Venus directly, or as a stand-alone program run right after the updated feed is published. The python code is trivial too. Doing this, subscribers using Google Reader or other advanced feed aggregators will get new content immediately, rather than on it's polling interval, whatever that is. >From a planet.fp.o as subscriber perspective, it's trickier. http://code.google.com/p/pubsubhubbub/wiki/SubscriberClients doesn't list any simple Python libs to "just do it". There is a plugin for the Tornado web server (which we don't currently use), and support for Drupal and others. This step might be "wait and see"... Thoughts? Thanks, Matt -- Matt Domsch Technology Strategist Dell | Office of the CTO _______________________________________________ infrastructure mailing list infrastructure@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/infrastructure
