See here.
NSF-sponsored papers should be freely available no more than 12 months after publication in a journal.
This is not perfect, but a step in the right direction. Computer scientists should insist that conference proceedings are treated the same way, and made freely available no more than 12 months after publication.
Hat tip: Lance Fortnow.
Windows On Theory 
awesome news, planning to blog on this also at some pt. fyi the link is 404. typo?
ok link works now. decided to blog on this also: big cyber/ network science milestones, arxiv/ NSF open access. fyi you guys dont have trackbacks turned on at this blog do you?