Some thoughts about putting research papers on the web

• Put your preprints on the web. I can understand not wanting to circulate unfinished papers (I have several I'm not yet ready to make public). However, if you're going to circulate a preprint widely, then there's no logical reason not to put it on the web.
• Keep your published papers on the web. If you publish papers but don't make them easily available, then it looks like you care more about racking up publication credits than about actually disseminating information.
• Putting papers on your home page is a good start, but it's much better to archive them somewhere where they will likely be maintained even after your home page is gone. For mathematics, physics, or computer science papers the arXiv is ideal. Greg Kuperberg has built a wonderful front end for the arXiv.
• You can link your home page to servers such as the arXiv, to save you the trouble of having to maintain your home page when a paper is updated. You can even link to all your papers automatically.
• Don't sign away your right to keep published papers in the arXiv. Why on earth would you sign over that right to the publishers for free, so they can turn around and sell your paper to libraries? That's anti-social.
• One of my pet peeves is notices on web pages that say the papers available there may differ from the published versions, with no explanation of how they might differ. In a couple of my papers I have fixed typos after publication (this is always noted in the arXiv comments), but otherwise every published paper is identical to the final draft I submitted for publication. (I suppose I can't guarantee that no changes were made by the publishers at the last minute, but I certainly did not authorize any such changes.) If you have out of date copies of papers on your web page, you should replace them with the final versions. If you actually signed an agreement promising not to do that, then it's an offensive agreement and you should never sign such a thing again.