This question arose out of a time with a lot of discussion on meta about how this site could be improved; it was proposed that we ban homework and this question explored how that would look like and whether people agreed to a concrete form or not. This question gathered a lot of votes (currently +25/-7 on the question, and 13/1, 11/1, 9/0, 7/0, 7/0 and 6/0 on the answers), but my impression is that there wasn't really enough consensus to implement this policy, or at least in this form.
Since then, the debate has slowed somewhat (as in e.g. this question) and it has drifted off into other directions (such as this question). In particular, next week moderators will leave all homework closures to the community as a way to try and find out what we, as a community, think should and should not be on this site. I regard this as a much healthier approach to the problem, and I think it should lead to this document becoming obsolete: we should reformulate the policy in terms of what we actually want and do, instead of some idealized construct.
To add a bit more food for thought to the meta table, I'd like to set down some retrospective thoughts on this proposal.
First of all, I must say I regard this as an experiment in writing and large parts of it as not very successful. It was proposed elsewhere that we ban homework questions but still keep the homework tag, at least temporarily. This policy proposal was an attempt to see whether such a policy could be realized in a concrete example, and I tried very hard to reconcile the two conflicting directions. I do not think I was very successful, and the policy below has several sticky points that must be resolved using judgement calls; I think this is a sign of fundamental contradictions at the heart of such a policy, and I'm doubtful that more successful implementations of the same core policy can be written while still retaining the homework tag.
The key point that sparked this was the claim/observation by David Z and Manishearth that homework is already essentially banned. I'm less convinced now that that's the case (see e.g. this search), but this was an exercise at seeing whether such an 'effective policy' could be set down into words.
Certain parts of this thread definitely still read nicely. I particularly like the emphasis on driving people away from homework-like questions and onto more conceptual ones, the shift to "what should I do if I want to ask a homework question?", and the introduction of sections on what to do if your question has been closed or put on hold or has close-votes on it.
Another contradiction is to tell people to drop the dependence of their question on the problem ("it should be useful to someone who's not staring at the same problem") but still telling them it is crucial that they 'reference the source' and provide the full context of their question. I'm not saying that we should drop either requirement, but I do think this merits a lot of thought as to how to reconcile the two once they're on (e)paper.
One useful practicality is that there are in fact a lot of small places where users with homework questions come into contact with the policy, which is reflected in the six different answers here. These should all be considered with a view on an eventual reformulation of formal policy.
I'll shut up now, as this page is already long enough. I just wanted to set these thoughts down, particularly since we're coming up on community closing. Any thoughts?