[edit by Anton: fixed hyperlink]

In particular, if one wishes to (as in fact I do) to be up-to-date with what is happening on arXiv, one can anyway subscribe to the daily-mailings of arXiv (in subject categories one is interested in). And, some people include in the meta-data of their preprints the explict request for comments. Of course, in that case the comments would not be public, but actually not sure if this is good or bad, yet if one really wishes to 'visibly' comment on an arXiv preprint one can: one just would write a blogpost, and there will be a trackback on arXiv (if I understand correctly).

as usual, we can't actually change the backend in any way; it's not our software.

That said, it seems at least plausible that we could explicitly allow "request for comments" questions on arXiv preprints. I suspect that this would in practice work very badly (in particular, the fraction of "requests" coming from the crank section of the arXiv would be unpleasantly high), but I'd be interested to at least have a discussion of the pros and cons.

Do you think that making such a site makes sense? Would the people behind mathoverflow be willing to do it? It seems to me that the changes to the mathoveflow backend should be minimal. Actually one could already use mathoveflow.net to post links to preprints if the rules allowed it.