I've been talking for years about how valuable a Journal of Null Results would be. Granted, it wouldn't be interesting necessarily, but a repository or database where one could look up someone's failed methods would prevent a lot of unnecessary reproduction of failed attempts.

Also, the ability to add comments (especially if authors' identities could be verified) to a research publication would be a revolution in peer-reviewed publications. If the PLoS is planning on adding that functionality, it will win many converts I believe.