My questions is whether there is a good reason to allow anonymous users in a site dedicated to research-level mathematics. In theory, most users of this site are professional mathematicians and graduate students. In practice, I'm not so sure. The amount of ill-posed questions, homework problems, or just plain nonsense that is periodically closed, but which nevertheless clutters the site, suggests there is a substantial number of other people who use MO.
In my view anonymity poses a number of problems:
We freely share ideas on this site and the mathematical conversations that take place in MO sometimes find their way to publications. It is one thing to say "the proof of this lemma was worked out in an MO conversation with name-lastname" than to say "the proof of this lemma was worked out in an MO conversation with Bullwinkle101". I actually have to do this in a paper I'm writing right now.
I have not done any serious statistics, but it seems to me bad questions have been driving the good ones out for some time.
It is easier to be nasty as an anonymous user.
Sometimes good mathematicians ask or say dumb things (as mathematicians, else the statement is truly obvious). There is no need to hide this. Actually, I think it is BAD to hide this. As a "guild" we believe too much in our own mythology.
If one takes a look at the first three or five pages of users ranked by "reputation" the number of anonymous users is propotionately very low and the proportion increases sharply as the reputation decreases. This suggests the question: what sort of users does MO wish to have?
Now, as a matter of principle, I think restricting access (to whatever) is not usually a good idea or, at the very least, has to be done intelligently. ArXiv has managed to do this nicely. In fact, if for a second we think of anonymous postings in ArXiv by whoever has an internet access and an interest for mathematics, we see that it could not be the useful research tool that it is today.
Thanks to all who answered this question or commented on it. I now see that there are some valid reasons to participate anonymously and that there are other or better ways to deal with some the issues raised in this question.