The TheoryOverflow site was proposed a few weeks ago for discussion of research-level questions in theoretical computer science. I am committed to using this service, and encourage others in the community to try it out.
Some of this kind of discussion currently happens online at MathOverflow. This means top mathematicians like Tim Gowers and Terence Tao are part of the discussion. But it also means some kinds of TCS questions are not on-topic, and many tend to be ignored by the majority of mathematicians who hang out at MO.
This also means that beautiful questions with a TCS flavour can languish with just a few votes. For instance, this question about computable models of ZFC set theory, which seems to have gathered a few votes from the mathematical foundations community, or this question about structural properties for Turing completeness, which generated a reasonable discussion but between a very small group of participants. Such questions seem currently to be mainly of interest to computer scientists, not mathematicians.
I am undecided about whether it is a good idea to reinforce the split between theoretical computer science and mathematics. Perhaps computer scientists could contribute more significantly to MO, as David Eppstein, Gil Kalai, or Ryan Williams have already done. But for now, TheoryOverflow is an interesting idea and I look forward to how it develops. I also encourage other theoretical computer scientists to commit to the beta. At the time of writing, 75% of the required amount of commitment has been generated.
I came to this conclusion after seeing how successful the recently created TeX, LaTeX and Friends site has been in just a few days of its active private beta period. (It has now opened to public beta, so please do join in!) Focused environments for expert-level commentary seem to be productive places, at least until trolls take over. This seems to apply to mailing lists, Usenet newsgroups, web forums, the communal tea room, conferences, or in fact any place people get together which satisfies certain criteria. (I’ll speculate on those criteria another time.)