There are many RSS sources on the web, and people are beginning to look at them as a form on unix pipe for the web. Whilst this isn't a bad analogy, it misses a number of things about pipes (especially graphline type pipe systems), which would make things more powerful. We've got a specific need here, so I'm thinking "what's the most Kamaelian way of doing this", and starting from the perspective of "what's missing". The rest of this post discusses the editted highlights of my thoughts.

What's wrong today with a web RSS feed as the universal pipe connector?

• It's pull only
• There's no apparent _standard_ way to push back - trackback is the closest I can find right now, but that's application specific, not a generic connector.
  
• There's no mechanism for actually creating the pipe aspect as yet today
• As well as pushing into the object you would then want to pull out the other sides (taking the Kamaelia viewpoint that you don't have just stdin/stdout)
• The web is not the only source. Any device could easily publish one or many RSS feeds via zeroconf technologies, meaning you could fire up your machine, find out who's publishing a resource locally, and then play with the resulting feeds.
  

Ironically on the Kamaelia project we've now got a specific need where this would be of use, so I'm extremely tempted to deal with this. As usual the specific need is trivial - posting subversion checkins as news onto the website, but it would be extremely useful. The resulting tools would be reusable, which could be quite fun in a number of different places :)