What are tag wikis for?

Question

Wait, wait, hear me out.

I was looking at subprocess earlier today. Its tag wiki reads, in its entirety:

The subprocess module includes replacements for functions such as os.popen*, os.spawn*, os.system, popen2.*, and commands.*.

This raises more questions than it answers, so I took a crack at rewriting it. Half an hour later, I realized I'd effectively been writing better documentation (the canon documentation is pretty bad) covering common facets of using this module, common pitfalls, etc. Is that what tag wikis are intended for? Am I stealing rep from authors of existing answers? Striving to be a good Research Assistant, in I dug.

• The sidebar says the tag wiki is "a detailed introduction to the topic", then immediately below, suggests it include a "brief introduction to the subject".

• The sidebar, and several meta posts, link to a post introducing the current tag pages. But it doesn't tell me what tag wikis are for, either: the advice at the bottom is a guide for writing excerpts.

• Tag wikis aren't searchable, and don't support deep linking. Thus, the more content is in a tag wiki, the less useful it becomes.

• scala is a "great tag wiki", but it's just an artificial outline, wishing questions were topic overviews. (e.g. "tail recursion", "List", and "non-nullable" describe using these features, not what they are.) Some items in the outline aren't even links, which raises questions instead of heading them off.

• This wiki functionality increase proposal cued discussion about what tag wikis are for, but has no resolution. If they only define the tag in the context of the site, why do we need more than a paragraph or two in the first place? Should the focus be on community wiki instead?

• Questions have answers, both have comments, all have voting. Wiki edits have only approval, and even that's gone once you have the rep. There's no real feedback on the most permanent prose on the site, which suggests that they should remain small and leave the real stuff to the answers.

Not exactly crystal clear.

---

Allow me to meander a bit. StackOverflow has accreted a remarkable amount of useful information, and definitely accomplished its original goal: searching for a specific problem is pretty likely to turn up existing answers on SO.

But general reference questions still feel like black sheep, even though "what is X feature?" and "how do I do general thing Y?" float at the top of many tags' FAQ. Nuggets frequently sneak into hard-to-find places. Meta still disagrees on whether artificial reference questions should exist at all.

Answering your own questions is encouraged, yet I don't know that I've ever seen it done. Should I take what I've written for the tag wiki and self-answer a question? Last year I started blogging a series of detailed Python FAQ articles based on topics I see come up a lot in #python. Useful fodder for SO, surely? Yet both "what is subprocess" and "how do I make a website" are awful questions, though seemingly ideal for a wiki page.

Shoehorning reference material into huge answers to vague questions preserves the format, but is artificial and self-defeating: they're bad dupe targets, they discourage other answers, they discourage edits if non-wiki, and they don't recognize the authors if wiki.

I see a big void/opportunity here. Wikipedia is terrible for expert topics, being designed for reporting on facts. The programming community needs best practices from people we trust. We're... not getting that.

---

I'm left, then, with two interrelated questions.

1. What does the ideal long-form tag wiki look like, today? Just a list of links to popular questions? Broad documentation when what exists is insufficient? Brief tutorials?

   Or from another perspective, who is expected to look at tag wikis? Do we have some idea whether askers think of the wiki as a potential resource to consult before asking a question, or is it just link-fodder for answers and comments?

2. Can we expand tag wikis to better address general topics? Clearly we want SO to contain all programming knowledge ever, but the software itself sometimes discourages contributing that knowledge. Fixing people is hard. Fix the software.

   We can't just have tag wikis with more pages. They need to gel well with the rest of the site and culture. Off the top of my head:

   • Answers should be able to easily quote chunks of the wiki inline (and add more context for the particular question, of course), and the author of that chunk should benefit from the answer's votes. Perhaps general-reference questions could be closed as exact duplicates of wiki pages themselves.
   • Anyone confused by part of the wiki should be able to ask a question about a chunk of the wiki itself, with an easy way to fold the answer (which may even take the form of an edit) back into the wiki.
   • Existing answers should be portable into a wiki without losing the context, the author, and the comments.
   • For that matter, chunks of the wiki ought to be individually commentable as well.
   • Wikis should be searchable and otherwise easier to find.
   • Wiki contributors should earn recognition—perhaps list the users who originally wrote the most content on any given page.

   Essentially, the wiki should expand the theme of answering questions by giving us a better way to answer broad questions and reuse parts of those answers. (I originally wrote my Python FAQ index as questions, yet I can't imagine posting them as they are to SO!) Community wiki is a good first step; tag wikis should be taking us further.

   I guess the real question here is: does this sound awesome, or am I crazy? :)

Answer 1

Stack Overflow is really good at creating information. But it's really terrible at organizing it so that people find what they're looking for. The organizational scheme essentially consists of "Let people search for it. And if our search system isn't good enough, let them use Google's". That's it.

SO is good if you're looking for an answer to a question that has been asked (generally. Depending on how well the OP named it, and how good the search engine feels that day). SO is much worse at another important task: finding general information.

If I'm having an acute problem, SO is the place to be; if nobody's asked the question before, I can ask it and get an answer. However, if I'm having a chronic problem (such as a general lack of knowledge about some programming subject matter), SO is the worst place to go. I don't know enough to ask appropriate questions. And learning from SO's existing questions is a terrible experience.

It's like sifting through random programming tests on some subject, hoping that some combination of random information may gel in your brain to form some mental image.

I think the idea behind tag wikis is that they can kind of bridge that gap. That is, if you're looking for general information on a subject, that's the place to go. The problems with this come from a number of factors:

1. Q&A sucks as a medium for general learning. It's a great organizer for someone looking for a specific answer to a specific problem. But for general information on a subject? It's very weak. And the fact is... some information simply can't really be presented as Q&A. Not without incredibly contrived questions.

2. Lack of formatting. I effectively maintain the OpenGL Wiki, and I've spent a lot of time writing articles and manicuring them into a reasonable form for maximum information distribution. And one thing I've noted is this: being able to control where information is is very important for navigability. For example, if I'm on the Vertex Shader page, I want to be able to go to the following:

   1. Since vertex shaders are a part of the OpenGL rendering pipeline, I need to be able to go to the previous pipeline stage and the next.

   2. Since vertex shaders are written in the OpenGL shading language, I should have a quick way to get to information about that.

3. Weakness of control. In my time on that wiki, I have on occasion needed to repeat information on multiple pages. And, like a good coder, I know this is a bad thing, for obvious reasons. But Mediawiki gives me a way to handle that: I can transclude from a page directly into the pages I want to use it. The only think you can do here is link to a new Q&A.

4. No information hierarchy. The founders hate the idea of tag hierarchies, but the fact is every programming book that has ever been written has been divided into information hierarchies. Parts, chapters, sections, sub-sections, etc. These are important organizational tools for cataloging information for a user to be able to learn from them. SO provides no mechanism for this.

Tag wikis don't provide any of this. All they can really be is just one page that links to questions (which presumably provide the actual info).

Q&A is for problem solving. Which is great if you have a problem that needs to be solved. But if you're interested in learning about a subject, it's just not helpful. And a simple tag wiki slapped on the side like a cancer isn't a solution.

Answer 2

I (as a 2 or so month participant here, but in Usenet since '85 and on many mailing lists) heartily agree. Q&A is fine as far as it goes, I am fully in agreement with "to the point" and other policies I see enforced here. But "short and to the point" doesn't help when looking for the wider picture. I have cited Wikipedia a few times for the wider context, and I've seen others do the same. But even the resources out there for the "bigger picture" fall short sometimes.

I saw discussions here, and some closed (!) questions (closed as "no real question", and with full reason) that are protected because they collect important answers (lists of C++ books is an example). Those perhaps should be collected into a FAQ, or set up a FAQ style area (perhaps a wiki? but with hard rules, the first post must pass extensive vetting, edits have to be approved with higher rules, answers/comments are allowed from the somewhat washed masses, hopefully somebody goes through the FAQs and integrates nuggets from later answers, or the best answers).

Wikipedia does try to cover much the same ground as SE, but from a different angle. I can only assume that there is a substantial overlap between the users/contributors at both sites. As contributors are a scarce resource, perhaps SE should encourage more collaboration (the licenses on contents are compatible, after all). Say if some area is well covered in the sense of a wider discussion as OP wrote, encourage the writer to contribute there and link to it officially from here.

[Perhaps need a "not a real answer" tag for this one...]