So we've had a bit of a debate in chat between the mods and some users (more users in chat would be awesome) about reference request posts. I thought, having slept on it for two nights, I would lay out the options as I see them, so here goes.
Firstly, declared bias: I am probably against reference questions on the balance of things (although I understand the desire for them). That said, it's not my opinion that matters, I'm only telling you so you know I'm biased!
The case for reference requests
- There is a growing community of very smart people here who know a good crypto book when they see one (or after they've read it).
- SE's voting system makes seeing majority opinion very easy.
- Reference requests are attractive resource for new users to the site. "Look, experts recommending books, I should read that".
- As questions on SO, book requests are mega voted. Our top questions are currently full of pretty good real questions.
- Other resources on the internet provide this feature. Amazon reviews, IACR's book reviews, personal blogs etc.
- Copycat questions, or "I want to ask about book reviews for XYZ". Overrunning the site with these, whilst it'll look good statistically, will look very poor quality wise and we now have excellent quality.
- Maintenance. SE employee Grace Note wrote this essay on gaming.se's meta, a site where game recommendations were frequently asked for. In short, there are problems with maintaining a "big list" that do not stretch to the question format we have. In terms of SE this means the question would regularly be bumped to the home page by edits ... or it will get forgotten (the answers get stale), which is even worse.
The possible solutions
Do not allow any kind of reference request at all. I don't think we actually have to do this anyway, because of point 2:
Use chat. We have a chat room and the rules there are much more relaxed. The chat room does not have to so rigidly on topic and between moderators and a couple of community members we have already used it once to talk about books. The only real rule of chat is "be nice". Also, did I mention we're running chatcasts (organised chat sessions)? One of these could include favourite introductory crypto reference? Yes?
Attach recommendations to tag wikis. Did you know we have tag wikis? For example, here's one (haha, chosen at random, get it?). You, that's right, you, can propose changes to fill these out (or edit them directly if you have enough rep) and there is nothing to say we cannot put recommended reading in there. Look at PHP or C++ on SO.
The question becomes how do we source any such book recommendations and is there a place we can provide generic, topic-introduction crypto books? We could use a meta question for the purpose of sourcing (or chat, as in point 2). To the latter, we're struggling perhaps maybe. I think tag wikis provide a nice niche way of topic focused books that are hard to find and fill the intermediate crypto resource problem.
We use a question on the main site. This is my least favourite solution by an awfully long way; I think for it to work there has to be one and only one such question and all resource recommendations need to go into it. By nature we would need to set the question as community wiki, which SE are not keen on at all. I do not think at this point in time we have any CW questions and I am not keen to set a precedent.
Aside from the problem with vote inflation, we also have runaway answers. After so many answers, a question actually has "pages" of answers underneath it, that's not really so good. It means you have to look through pages. How many of you like reading page 40 of a phpBB thread? Not me. It means book recommendations on page 40 are not likely to be seen, meaning they won't be upvoted, defeating the purpose of using the voting system. There are not any real practical ways we can enforce this. Keeping the answer pages short means combining book/answer into topics, again reducing the point of voting to precisely 0.
Finally, I mentioned locking in chat, which would prevent vote inflation on the question and additional answers being posted when applied. The purpose of locking is to (ideally temporarily) prevent edit wars on a post and vandalism where no other technique works. Applied to a post, nothing useful can be done to it (vote, edit, comment, delete, close). Applied to a question, this locks down the question but answers can still be edited/voted/commented.
Using this tool in this way would represent an abuse of the mod powers in the view of SE, I expect, and I agree with them. It takes away the ability for the community to maintain the resource and as moderators we just shouldn't be doing that. The tool exists to prevent and stop abuse and should be used only for that purpose. To date, no post has been locked on crypto, ever (other than the stubs for migrated and merged questions, which automatically get this status), and in spite of discussing the idea, I think my viewpoint is now that we should not do this.
There is a question in here and here it is - can we as a community agree a strategy for handling reference requests? And the main point I'm trying to ask - does this really need to be in a question/questions? Or can we use the various other tools we have (tag wikis, chat) to satisfy people's needs?