One of my buddies (michael) is doing something similar with annotations and social reading @ pen.fm. I'll see how we may be able to incorporate some of these ideas.

I do have around 450k papers from JSTOR (released legally via archive.org). I have these papers indexed and can expose them via openjournal.

Also, as I mentioned in my previous post, a few guys from Berkeley (Tony Chen et al) are work on peer library which will be a more complete non profit academic search engine.

Good points about the 'barrier'. I'm hoping people will contribute their own papers (even if they are no published).

My ultimate vision would be for people to write their papers within a git repo and then upload/submit their .tex source (along with unit tests). I'm in the process of building some of these features.

Also, experimenting with some ocr and pdf analysis to scrape as much useful contextual information as I can from the papers contributed (for the great benefit of our users)

Thanks for taking the time to respond, streptomycin!