also specific forum discussions and wikis may be useful: my last arxiv preprint has been essentially developed on http://www.math.ntnu.no/~stacey/Mathforge/nForum/ before being put in paper form. It may be of interest to this topic that teh three authors of http://arxiv.org/abs/1011.4735 have never meet face-to-face and the whole developement of the paper has been via web tools.
(sorry for making an example in which I'm personally involved)
