We got our first list question. We need a policy how to deal with this and similar questions.
Real questions have answers, not items. Reading lists don't work well in the questions and answers format of Stack Exchange. With one item per answer, you end up with voting that combines how old an answer is (utterly useless), how popular it is (mostly useless), and how good and relevant it is (if you're lucky). Your proposal removes the impact of age, but it requires heavy duty maintenance: we'd have to watch every list question, check how the voting goes; and in the meantime we have this jumble of answers (many of which will be in the provisional stage for a very long time).
For book lists covering a topic, we have a place: the tag wiki. There is a Stack Exchange-wide move to put book lists in tag wikis, and better support for large tag wikis is in the plans.
Other book questions may be acceptable, but only if they call for useful answers. That is:
- The question should express a specific need. Not “all the books/the best books on topic X”, but “I want to study X, what combination of books/articles would give me good coverage?” — where X is narrower than a whole sub-discipline.
- Answers should strive for completeness: never give a single item unless you claim that this single item is sufficient in and of itself. An answer that adds just one book as a complement should be an edit to an existing answer or a comment.
I recommend reading the Literature Stack Exchange book recommendation policy (which kinda works, but still lets a lot of mediocre questions with mediocre answers in) and the Programmers Stack Exchange book recommendation policy (which… well, it's Prog.SE).
Community wiki is not an excuse for bad questions. If the question is good, it doesn't need CW. If the question is bad, CW won't redeem it, it needs to be closed.
We don't need to tag these questions specially. If the question is good, it doesn't need to be ghettoized. If the question is bad, a tag won't redeem it.
Regarding this specific question, I initially voted to close, but I've halfway reconsidered. I think it may be barely ok. The good thing is that it calls for a limited syllabus (we're not going to recommend 100 books). On the other hand there are already 6 answers, of which only 1½ really address the question; the other answers just take this opportunity to mention their favorite books. Furthermore, the lack of a tag (no, soft-question is not a suitable tag) is a bad sign.
In the light of this discussion, this is what I propose to do:
- Ask users to propose only one book per answer.
Rationale: Votes should reflect wether books should go into the main answer. Having only one book per answer simplifies that.
- Create one CW answer (per field (algorithms, mathematical foundations, logics, languages & automata) if the list becomes long).
Rationale: CW because there should be no credit where none is due.
- Incorporate all answers with score at least five and ask the users to remove their answers.
Rationale: One central list with good books is better than redundant, scattered information. Removing incorporated answers keeps the thread clean.
- Remove answers that do not reach score 5 after two weeks.
Rationale: Keeping the thread clean.
After a while, i.e. when the list has converged, I think we should consider protecting the question. That is an easy way to prevent unhelpful late answers to cause work. Further additions can always be proposed and discussed in chat.
What do you think?