While a person can become a pro in one topic, it is not possible to master all the fields; even if we narrow it to computer science. Interdisciplinary fields make the problem worse. For instance, someone who works in (theoretical, not applied) cryptography must have a good knowledge of the computational complexity.
What happens when you need to ask a question which, while related to your research topic, happens to be lie within another field? Well, a complicated problem for you might be too basic for researches in the target field.
I like the model of Math.SE / MathOverflow a lot: If I think my question is too basic (which usually is!), I'll ask it on Math.SE. Research-level questions can be asked on the latter.
This discussion has recently gained momentum on Area51; here are some links:
And even proposals like Beginner Theoretical Computer Science or Computer Science were considered; but they were soon deleted/closed due to lack of support / misunderstanding that they are exact duplicates of other SE sites.
Lack of Support: I believe that such proposals did not advertise themselves well. For instance, if I ever knew that they exist, I'd join them at once. This is the main reason I'm raising the issue here.
Exact Duplicates: Some of the voters (to close) claimed that such proposals are exact duplicates of StackOverflow, while others claim that they are exact duplicates of TCS.SE.
- The former group believed that anything in computer science which is not theoretical is programming related. In my opinion, that is plain wrong. A basic question about Turing machines is a simple counterexample.
- The latter group were unable to distinguish between research-level questions and basic ones.
Some discussions can be found here: Is Computer Science a duplicate of Theoretical Computer Science.
In an extreme case, Jonik suggested: "Instead of another site, how about modifying the description of the existing site to allow a little less advanced CS theory questions too." His suggestion was unwelcome, and received some down-votes.
In another thread, Joel Spolsky requested that we (TCS.SE) broaden our view. His suggestion didn't go anywhere either.
The bottom line: I believe that CSTheory must have a non-research-level counterpart. I ask people here to help (re)create one on Area51, and support it adequately.
PS: There's always an issue of homeworks: Professors don't like their students to cheat by asking homework problems in online forums. This can be handled as in Math.SE, or by developing adequate policies at the new site.