But which of my semantics papers should be posted? We get asked this a lot. We can immediately offer two clear guidelines: Don't be shy! People want to read your papers--that's what we do. In particular, we are eager to have posts not only from senior researchers, but also from graduate students. Second: don't worry about taking up space on the server. For now at least, we've got lots of room. We encourage you to post multiple versions (e.g., multiple formats) of works of any length.
But ultimately, this decision is up to you. Posting should satisfy the following three criteria: first, the paper should add value to the archive. If you can imagine a semanticist downloading your paper, it adds value. Clearly, old manuscripts as well as new ones add value. Second, posting the paper must be helpful to you. If you want to increase the readership of your work, then consider posting. Third, posting the paper must not violate copyright.
This third point is the tricky one. Strict legal rules do not currently quite match commonly accepted practice in Linguistics and other fields, and there are some difficult cases where conflicting interests must be balanced. On the one hand, if a journal owns the copyright, they have the legal right to decide where and how the paper is made available to the public. On the other hand, if it takes two years or more for a paper to go through the refereeing and publication process, you, the journal, and the field will all benefit from making a pre-print version available sooner: wider exposure generates feedback, which improves the quality of the revised paper and generates interest in the published version. Will anyone cancel their subscription to L&P because of the archive? We strongly doubt it.
We're not lawyers, of course, but our rule of thumb is: if it's appropriate to post a paper on your publicly-accessible professional web site, it's appropriate to post it on the archive. Think of the archive as a meta-web site that gathers in one place some portion of individual semanticists' web pages. Also, bear in mind that you can always delete an item when it finally becomes available in print.
We recommend seeing how other archives handle this question. See, e.g., the policies at the Rutgers Optimality Archive, or the lack of a clear policy at arXiv.org, one of the first, most prominent, and heavily used pre-print archive for the hard sciences.