General Reference questions create low-quality dead ends for Googlers

Question

Introducing a "general reference" close reason was rejected by Jeff and Joel for fears of abuse. Jeff said in a comment:

Joel is very concerned it will be abused. I have concerns as well, since many programmers are of the "cannot abide a single atom of duplication in the omniverse" OCD variety

However, the bigger Stack Overflow grows, the more these trivial questions with their highly-upvoted trivial answers are actually making the Internet a slightly worse place by providing outdated information. Today, a discussion brought to light a nice example of the phenomenon:

How do I get the last inserted ID of a MySQL table in PHP?

Here's what's wrong with this question and its many brethren:

• It's a trivial question.

• It has plenty of wrong answers, but that's just an aside.

• The highly-upvoted currently accepted answer is absolutely correct, but it is referring to an outdated database library, while the current recommendation is to use the PDO database layer. Fixing the answer is of questionable value, as the outdated answer is still correct in the context the question was asked in.

• It's textbook "general reference". Googling the question title PHP: how to get last inserted ID of a table? leads you to the canonical references in the manual for both the outdated, and the new recommended method.

• The answers contain no mention of those canonical references

• so the whole thing is a low-quality dead end for the incoming Googler.

• It has over 59,000 views!

Having these kinds of questions around sucks, and is in the worst case actively preventing people from finding the correct canonical info by occupying slots in the Google results.

What should be done about these kinds of questions?

• Can "General Reference" be revisited?

• Can a protocol be developed to burninate these kinds of questions altogether, or to link to the proper documentation instead? Because that's the only thing that would make sense IMO.

• Can something else be done about this? What?

Answer 1

It seems to me that the problem is not with the questions, but with the answers. What's a trivial question to you is not a trivial question for a beginner. Indeed, an accepted answer which consists solely of this isn't very good:

Use mysql_insert_id()

Even then it could be worse, at least there was a link to the actual documentation.

deceze's edit significantly improves the question. Multiple possibilities, with a discussion of when they should be used: this is how a Stack Exchange answer should look like. Stack Exchange is, after all, a questions and answers site, not a link collection.

Do note that your example question is not a “general reference” question as generally practiced on the Stack Exchange sites that have it (Science Fiction & Fantasy and English Language & Usage). “General reference” started out as

this question is too basic; the answer is indexed in any number of general internet reference sources designed specifically to find that type of information.

and was made popular via Borror0's diagram. But over time I've come to like this diagram less and less, and so have many other participants on SF&F. I'll reiterate what I wrote before SF&F even launched:

It's “Wikipede it first”, not “Google it first”.

(Substitute Wikipedia for whatever reference is applicable: a dictionary, the library reference manual, …)

Relying on Google is problematic, because it doesn't say anything about how reliable the search results are. If you say “google it first”, you may be refering people to W3schools — are you sure this is what you want? If I google the title of your example question, I get:

1. Stack Exchange
2. Stack Exchange (duplicate? I'll let topic experts judge.)
3. official manual — mysql_insert_id
4. official manual — mysqli::$insert_id
5. Stack Exchange
6. Coding Forums
7. W3Schools
8. Vision.to
9. Hardcode
10. PHP Rockers

I have no idea which of the results are good answers. Keep in mind that the official PHP documentation may have its page rank boosted by being linked so much from SO, which has a huge rank. If you suppress this kind of question on SO, the first search results may well be on sites that peddle bad advice.

On SF&F, most of us have come to agree that Google is irrelevant in determining whether a question is general reference. We only close questions as GR when the subject matter can easily be looked up in a known-reliable (or at least presumed-reliable) reference — for us, this is mainly Wikipedia. I don't have the exact number, but not many questions have been closed as GR, and several of those involved debates as to whether the question was really GR or whether the Wikipedia article had enough detail — and if the question triggers such debate, there's a good change that the subject is not a “too basic” one. In a recent debate, we concluded (+20/-5) that we could easily live without the general reference close reason, because it causes more trouble than it's worth.

I'm a moderator on both SF&F and one of the language sites (not English, which has the “general reference” close reason, but French, which doesn't). My experience on SF&F is that GR isn't really useful as a close reason. Ok, we can close a few questions that aren't helping the site, because Wikipedia answers them better. But it's not much in the grand scheme of things — and there is an argument that GR questions are “getting-to-know-you” questions that drive traffic to the site.

On language sites, I think GR does more good than bad. We do get questions that are not doing any more than asking the meaning of a word, and where the asker is satisfied by a dictionary quote. Even then, I have seen GR being abused, when the question was about fine usage distinctions that are not covered in dictionaries. Still, I think GR would be a useful tool on language sites, as long as (like other tools) it's wielded carefully.

By the way, since French.SE doesn't have GR as a close reason, we use “too localized”, because a question that can be looked up in a dictionary is not going to help future visitors — not when they would have an easier time looking up the answer in their dictionary than finding it via a search engine or Stack Exchange search.

On Stack Overflow, I think GR as a close reason would be awfully abused. Programmers tend to say RTFM a lot. Sometimes it's warranted, sometimes it isn't. GR closure contributes to drive the quality up by excluding too-basic questions, but it also contributes to drive the quality down by excluding interesting answers. (Think Eric Lippert.) SO already struggles to provide answers that go beyond providing a line of code and explain how and why it works. GR closures would make it even harder to get answers to “why” questions.

We're circling back to my original point: basic questions are fine if they attract interesting answers. If you feel that a question truly does not merit a better answer than what is in the official manual, close as “too localized”. If you feel that a question is basic and boring, downvote it and move on. If you see a basic question with a one-line zero-effort answer, do not upvote it and write a better one.

The real problem isn't basic questions, but basic questions with basic answers. You can fight basic answers by not upvoting them. Or by dispelling the fiction that more upvotes means a better post. Yes, I realize this is hard — most upvoters aren't reading Meta, and that big number next to the answer makes it look good, as does the big number next to the author's name. But it's the way Stack Overflow works: quick and dirty answers are favored. Encourage long, thoughtful answers, and the good content will naturally overshadow the too-basic content.

Answer 2

There are a few things you can do to help reduce this problem:

1. Improve the title.
   Many questions like this have titles like "Using Foo to Bar the Baz." These titles show up in Google Search results, and when you navigate to the post from Google, you find out that they're actually asking about an error, and it doesn't have the information you need at all.

   Take a few moments to provide a more meaningful title like "InvalidFrobError error occurred while using Foo to Bar the Baz", so that the next person reaching the question from Google doesn't have to endure the same frustration.

2. Provide a link to a canonical resource.
   Link to the General Reference in a comment below the question. If the canonical resource is a Stack Overflow question that adequately solves the OP's problem, close as a duplicate. If it's an external resource, close as Not a Real Question.

3. Cultivate a collection of canonical posts
   The PHP, C and C++ tags do this to great effect. When someone asks a very common question, find the best post that asks and answers this question, and close all others as duplicates. Make a list of links to these "canonical posts" in the Tag Wiki.

Answer 3

One way of adding this while stemming the potential abuse (at least partially) would be to provide a link for the canonical reference. That way when questions are posted that are basically just asking something that should be in the docs anyway, we can close it "against the docs". So the canonical reference point would be the docs instead of as a duplicate question on SO.

Answer 4

I was the biggest proponent of introducing General Reference not that long after I was elected, for precisely the same reasons that you describe. I spent a long time digging through the data explorer and the site looking at examples where the reason would be applicable given the criteria of 'general reference':

General reference: this question is too basic; the answer is indexed in any number of general internet reference sources designed specifically to find that type of information.

I was in touch with Jeff repeatedly while doing my digging, in fact I traded several e-mails with him prior to spending any real time on it just to see how the idea would be received. Jeff wasn't exactly 'all for' the idea, he just wasn't as against the idea as Joel.

There was a podcast (I can't find it at the moment, weird) and Joel continued to make the same point that I began to believe myself while looking at examples where the close reason would apply. People would misuse the reason to close anything they found to be too basic. Even if I had been on the podcast while Jeff argued with Joel to a degree, the examples I had just weren't strong enough to argue, and there weren't many better examples to try again with.

During the hours that I spent going through both the site and data explorer, I expected to find plenty of questions like this:

How do I write a loop in C?

or

How do I define a variable in Python?

Yet, those didn't exist, at least not in enough frequency or consistency to back up an argument that the close reason is needed and would be helpful. Most of the examples that I could find where it would apply were things like:

How do I (do something) to a string in (language)

While easily solved by Googling, these questions are answerable and do demonstrate at least a modicum of prior programming knowledge on the part of the author (barely enough to satisfy the 'programming enthusiast' criteria, but still just enough).

In short, and tl;dr: - We're extremely unlikely to ever see General Reference added as a criteria for closing. Canonical posts are the best way to deal with this, but they can be a bit of a pain to find depending on how the question is asked.

I think what we should be looking at instead is ways to improve the tag wiki system to help deal with these questions and better organize the canonical gems, as well as the many ways a single basic question can be asked.

Answer 5

If we do end up with a General Reference close reason, then it should put a link at the head of the question, like the Exact Duplicate close reason uses now, pointing to a canonical question/answer.

Answer 6

You've done a good job of reiterating the problem here, but not of addressing the concerns over a new close reason as the proposed solution. Keep in mind, the questions you're explicitly calling out:

• Identify (or at least appear to identify) a common problem. These aren't obscure questions that no one will miss - chances are, closing them just introduces a new game of "whack a mole" into the system, annoying hundreds - potentially thousands - of folks sharing that question in the process.

• Are, by the normal rules of the system, valid questions. There's a real problem here, with a real solution. It's not subjective, argumentative, overly-broad, ridiculously narrow, or poorly-specified. You can't tell someone they should have asked it differently, you can merely say "you should've googled it - or just gone to [insert canonical site here]". You're effectively asking for the reversal of the long-standing policy against "just google that" answers.

• Are only annoying to folks who already know the answers. Now, I'm not saying an incorrect answer won't annoy anyone reading - but there's no good way to identify a general-reference question that'll get an incorrect (or soon-to-be obsoleted) answer when it is asked - indeed, the criteria for Stack Overflow would likely be rather different from the criteria on a site like English Language & Usage or SciFi & Fantasy, where "General Reference" means "use a dictionary rather than asking us to re-invent the wheel" - on SO, there may well not be any reliable, canonical documentation when the question is first asked.

In short, what I think you're getting at here isn't the need for a General Reference close reason, but something more akin to "Close as Obsolete". Which then raises the question: how would we define such a close reason?

---

As an aside, these sorts of questions are one reason why I dislike some of the witch-hunting surrounding "link-only answers". As originally written, your example's top answer was extremely bare-bones - it consisted of a terse link to the documentation which was updated by the maintainer with a big "DEPRECATED" warning right at the top and a link to the preferred API! Even when it became out-of-date, because the answer linked to a canonical source it quite possibly remained valuable (and now that it's been edited, it should remain valuable). Not only does this provide a "signpost" on The Internet for folks who aren't aware of the canon, it can even improve the ranking of the canonical source in future Google searches!