Anyone made any progress on this front?
For me the ideal tool would integrate with
CiteULike. You should be able to synchronise your pdf library with the server (although there would be some issues with annotations) and keep a synchronised bibtex library that you can browse and maybe edit on the iliad. The killer features would be pushing back to CiteULike those papers that you have actually read and when you read them as well as an OCR'd version of your annotations as a private note.
The more I think about it the less work this seems (apart from the OCR which isn't really needed, it's mostly just a bit of logging, synchronising files and scripting some HTML gets and puts - it could mostly be done on a PC and then just copied back and forth) and the more useful it could be...
But I don't even have an iliad yet, never mind a working knowledge of what is and isn't possible on it, so this is all a bit pie in the sky...