nodalpoint.org

I get my standard Pubmed searches in RSS through Hubmed. After searching, the orange "Feeds" button in the uppper righthand corner will take you to the appropriate pages; once you drop the RSS feed address into your aggregator (are they still called that, or is there a nom du jour?), it will poll Hubmed and update your search as appropriate. You can, of course, also generate the query URL on the fly.

One limitation of this atm is that only the first 20 hits are returned, so although you build up a comprehensive search feed over time, it's effectively limited by the date you start. A workaround I've discussed with Alf, the brains behind Hubmed, is to use the Pubmed Eutils to retrieve the original query as XML, which you can then convert to RSS with, for instance, an XSLT stylesheet or perl script. Given I know nothing about RSS syntax though, I haven't done this yet. Any gurus in the wings willing to take the challenge?

Yup, CompleteRSS is pretty complete.

I guess by "relevant and comprehensive" I was thinking journal contents. I agree that as contents update infrequently, a lot of feeds just sit in the reader for days/weeks, but personally I find this preferable to sifting through my inbox contents. IngentaConnect is not my cup of tea (fortunately our uni actually appears to be subscribed).

As for anything Yahoo, I suspect you won't get much sympathy for that point of view round here! If you want RSS on a website and run a server, I quite like Booby, a sort of online PIM. Evolution is also an option, though I don't like Evolution much (mainly because I like my email separate from other things).