Chris: There's another peculiarity in Perelman's way of sharing his results. As you note, he's posted his three papers to arXiv and benefited from its form of open review. But to date he's refused to publish the same papers in peer-reviewed journals. The problem is not that a conventional journal would refuse them, at least now that consensus is building that his proof is sound. Steven Krantz, editor of The Journal of Geometric Analysis, has offered to publish Perelman's three papers or any new ones that he would like to submit, but Perelman has not accepted the offer. (See my blog postings on this here and here.)

Perelman is not just saying, by his actions, that open review suffices. If that were all, he could still accept conventional publication for whatever increment of authority it would bring e.g. for mathematicians who are waiting for the imprimatur of a peer-reviewed journal. He's apparently saying that there is no such increment of authority and that he doesn't care to persuade mathematicians who think there is.

PLoS ONE uses two forms of review --one closed, internal, and prospective and the other open, external, and retroactive. I believe that Perelman would only support half this model.