A syndication formatting cache

I'm really thinking about this stuff just now, so this note is as much use for me as anyone else.

We have a ton of sources all working with Atom/RSS formats but being semantically different and in cases extending the same concepts in different ways (e.g. Digg has it's own namespace in its Atom feeds for authors).

Imagine a service that indexed and transformed these sources to normalized formats. So you could basically do XPATH style queries (the interface wouldn't be so complex of course) on the RSS/Atom sources and not only get the data in a given element, but be semantically accurate on what you are getting.

In addition, extension namespaces could also be queried, so you could ask for media items from youtube, flickr, meefedia and so on and get an accurate result.

This service may even be useful if we were all using the same format, but at a time where joining feeds is near impossible when you are thinking about the user, it may be useful to have a service helping out.

I've already written a bunch of tranforms to do this and to be honest had to write an Xslt for every single feed i got (i think delicious was ok), so i know the headache as others look for more advanced syndication feeds!