Update: according to http://tea.mathoverflow.net/discussion/67/mathoverflow-irc-channel/?Focus=703 it's on freenode.

One may argue that a forum has even better facilities (for instance, on a forum you can quote other people) but on the other hand on a forum there are fixed categories. On StackExchange we have multiple tags for each thread, which allow a finer classification.

Nope. Forums don't have to have fixed categories. On the nForum, registered users can create their own categories if they don't like the ones that I've provided. Moreover, there are extensions to the software running that forum (and this) which allow for tags (I just haven't installed it on the nForum as it hasn't been asked for). There are also extensions that allow for community moderation and for voting on posts. So actually, while it's a lot of work to get SO to act like a forum, it's not that hard to get a forum to act like SO, so if you really want discussions, I suggest using a forum software.

Regardless of whether SE or MO are good for discussions, some people will believe it is, even if only temporarily. Should we do anything more than we're already doing to prevent people from trying to start discussion questions? If yes, what?

I think the current approach of simply cultivating an atmosphere where discussion questions are discouraged is pretty good. I can't think of any software features that would help. Then again, maybe I'm just unimaginative. Any ideas?

If I post a comment in reply to Konrad's answer, Konrad gets a notification but Theo doesn't. If Konrad then posts a reply to that, neither myself nor Theo get notified! Plus, how do we cope with the limitations on formatting in comments? So maybe we should post answers instead, but then they aren't sorted and also make it confusing to someone interested in the answer to the original question.

The main point of my post, which noone addressed, is whether we can formulate a sensible feature request for StackExchange which would favor precise questions with respect to discussions. In particular I don't think a question should bump on top on each reply; instead I'd like to see some adjustment in such a way that questions with few answers stay on top more than the others.

regarding the skew of subjects on the arxiv: I've heard actually that there is still a significant skew, matching the historical origins of the mathematics arxiv(s). There are some fun statistics at http://arxiv.org/Stats/hcamonthly.html, but I'm not sure there's enough there to see what's going on in this regard.

• meta.SE: Does Stack Exchange suits a discussion site? (forum)
• why close questions? (this post actually explains why we wouldn't want discussions even if SE were good for them)
• Appropriateness of Questions
• Philosophical Questions
• Soft Questions on MathOverflow (I think this is one of the discussions where I'm trying to bring Andrew over to the dark side)

@Andrew: If you don't mind my asking, what changed your mind? Was it discussion here on meta (if so, please link to the relevant threads if you can find them), or was it using MO itself?

I dislike the idea that MO should become "All things to all mathematicians". It took me a while to come round to this, but I encourage anyone who thinks like this to read some of the discussions that Anton and I had here on meta to see what led me to change my mind.

A bit later, I changed my mind: see http://tea.mathoverflow.net/discussion/210/close-cocktail-party-math-and-fourier-transform-for-dummies-and-assorted-comments

I agree that the SE software would probably be reasonable for discussions. But I don't think that's the point. The point is that MO is meant for answering well-posed math questions. I think that if we allowed discussions, we'd run the risk of allowing the site to be overtaken by discussions and we'd possibly compromise the question/answer aspect of the site.

I don't agree with your comment:

Whenever a question has a really precise answer that only a specialist can give, most people just don't reply and the question soon fades into the second page and becomes forgotten.

I don't think it gets forgotten. At the very least, the asker remembers it, the answerers remember it, and Google remembers it. And the next time someone in the world has that same question, they'll be able to find the answer.

This site is generally pretty good for hard questions in algebraic geometry, but I've noticed that hard questions in other fields very often are left unanswered and with a low score. Mediocre questions in algebraic geometry often end up with more votes than much better posts from any given field (and I say this having asked a few mediocre questions in algebraic geometry =p). This effect combined with the number of popular soft-questios/big-lists very often makes it very frustrating to ask anything here at all. The first problem seems like it could be addressed by closing down discussion and big-list questions by moderators, but I'm at a loss on how to deal with the problem I mentioned.

I'm starting to wonder whether this is actually true. I should make clear that I like the role that MathOverflow has for math questions, and I don't like very much long and vague discussions. Still, I think that StackExchange is quite a good platform for discussions and less so for questions with a definite answer.

1) StackExchange is a good software for discussions. Indeed it is very easy to follow discussions here. We have favorite questions, feedback whenever someone anwsers or comments our questions and answers, and more important, every time there is an answer, the question is bumped on the top of the main page. All these tools make life easy to follow discussions, certainly more than a blog where everything is linear and long discussions immediately become cumbersome. One may argue that a forum has even better facilities (for instance, on a forum you can quote other people) but on the other hand on a forum there are fixed categories. On StackExchange we have multiple tags for each thread, which allow a finer classification.

2) StackExchange is not a very good software for precise questions with definite answers. Of course, this is not literally true, as both MathOverflow and StackOverflow (at least) show. Still, it is annoying that discussions always get bumped at the top, and it seems that the only reason why the first page is not all about growing children or mathematical jokes is that moderators are patiently keeping track of these discussions and closing them after a while. Whenever a question has a really precise answer that only a specialist can give, most people just don't reply and the question soon fades into the second page and becomes forgotten.

So I think we are really cheating when we write this standard advice. StackExchange IS good for discussions, as it is witnessed by the many active discussions on MathOverflow itself. So, what is the point of this post?

For one, I think that it would be more honest to write that discussions are simply not welcome here. I do not see any technical issue with having discussions, other than flooding MathOverflow and hiding the most interesting math questions.

More concretely, I wonder whether we can make some sensible feature request for StackExchange. For instance, one may want to assign to every question a weight which is more or less inversely proportional to the number of answers already received. The highest the weight, the more plausible you will find the question on top of the list. One may also want to take care of whether the question has an accepted answer.

In this way questions with a few answers will be bumped often, and hopefully get eventually answered. This would discourage big-list questions and help those still waiting for answers. Note that this would be more effective than periodically bumping unanswered questions, since a question may very well have just one or two answers which are irrelevant.