We have a great influx of questions that are permutations of the same basic question. Among the most common examples in the tags I'm active in are

  • How do I parse this HTML structure with PHP?

  • How do I create pretty URLs like www.example.com/name/number with .htaccess?

sometimes with a dozen new questions in one day.

It is my opinion that most of these questions - those lacking any kind of originality - should be closed as duplicates, or closed in some other way. (I asked a Meta question about this, but got little feedback.)

I started a question titled "how to parse HTML with PHP?" a few days back. The intention was to build a reference question that contains the best generic solution to this common issue, with the goal of having one "best" option when it comes to closing as a duplicate.

It's CW, but I've promised bounties for the best answers to create an incentive to contribute. (I'm planning to leave the programming business, and thus cease daily participation in SO, next year. While I may keep my SO account, I want to bring my rep points to zero. My favourite idea would be to spend it on something useful like this. If this specific idea is rejected, it's no big deal either.)

The question has been relatively successful in terms of votes and responses, but has received very good points of criticism as well. Most notably from @Gordon, who by the way is a very active duplicate hunter.

Closevoting this, because IMO a reference question will shift attention away from the actual original answers towards this one, lowering the chance for reputation gain on the original answers. If anything, a reference question should collect the major topics and link to appropriate answers, so people still get a chance to be upvoted. It's more like a digest thing for those too lazy to use the search function then. How to approach FAQ question should be elaborated on meta first.

I can see a point in this. I also realize that creating "reference questions" like this is a bit artificial: It goes against the natural flow of people asking real questions, those questions getting answered, and the best questions and answers bubbling to the top through upvotes.

It is my opinion, however, that a solution for this is needed, and that having a reference question is a good thing. When I'm voting to close something as a duplicate, I want to make sure I point to the right duplicate, one that really helps the OP. Usually, I will look into the highest voted questions first to find some kind of reference. I love this way of finding great content. What I'm proposing (and what I would like a community ruling on) is formalizing this, so that closing duplicates becomes easier.

My questions are:

  • Are "reference questions" like this a good idea at all? Should the experiment go on, or should I close it down?

  • If yes: Is the execution okay? (Entirely new question, made CW, bounties for the best answers....)? Or should, instead of building a question from scratch, the best (most viewed / most highest-voted / most answered) existing similar question be used? (Every issue that needs a reference question already has brilliant answers somewhere.)

  • Or should we let the question base grow naturally, and view the plentitude of similar but individually specific questions as a compendium where, over time, everyone can find the solution to their problem? Although I don't share it, because askers seldom look for duplicates, it is a valid view.

  • Are there other ideas to deal with the issue?

  • 3
    Here's a reference question Gordon just posted. That's IMHO tag wiki material, but NaRQ. I also find it hypocritical to post this after closing your question. Sep 17, 2010 at 16:39
  • @Null it is in the format he suggested we use instead, so I think it's a fair shot (and needs the community's decision whether it's the way to go). The Tag Wiki aspect is a good point. I have played around with boosting an existing question to "reference question" status using a bounty but that's a pretty expensive way. Interested to see how the discussion plays out. It is definitely a good one to have
    – Pekka
    Sep 17, 2010 at 17:30
  • 1
    I think the minimum rep too is 1, so you won't be able to go to all the way to 0.
    – alex
    Sep 30, 2010 at 12:39
  • The moderator note needs to be removed from the question if this is outdated.
    – Incognito
    Sep 21, 2011 at 20:08
  • I also thought about this. Nice to see that it got such a good feedback from the community. Can you or anybody provide a list of currently existing reference questions?
    – hek2mgl
    Jun 5, 2014 at 18:23

10 Answers 10


I think reference posts are a great idea, like the ones that polygenelubricants is doing (mentioned here).

They need to be very specific and focused, with just one basic concept being addressed.


I see many SQL questions of the form:

  • How do I get the most recent order for each customer?
  • How do I get the record with the best exam result for each student?
  • How do I get the game with the least score for each player?
  • etc...

They are all effectively the same question:

  • How do I get the row with the greatest/least x for each group?

Unfortunately it is not a trivial question to answer and often slightly different approaches are given each time depending on who happened to be online (different people have different preferred approaches).

They are usually tagged as greatest-n-per-group. This makes it easy to search for other similar questions and re-use their answers. But I'd rather just close it as an exact duplicate. The problem is that it never is an exact duplicate because the wording is slightly different and the table names are different. Maybe even one time there's a slight twist but the basic principle is still the same.

If we could have a reference question for this and let the OP do the appropriate substitutions to make it work for his specific table names then it would save everyone else a lot of time repeating basically the same thing over and over. We could also have a place to discuss which of the many approaches is the best, or in what situations one approach should be preferred to another. This question, if repeatedly linked to, would get more attention than all the other questions so hopefully the quality of the answers there would become high.

So I think it's an interesting idea. Currently I just basically repost the same answer with appropriate search/replace on table/column names and other minor adjustments. I also sometimes link to the 'greatest-n-per-group' tag so that the OP can search for other alternative solutions themself. But linking to a reference question would perhaps be a better way to handle this type of "almost-exact-duplicate" question.


I can see how "reference questions" could potentially help the site but I'm not sure it's the best way for StackOverflow. StackOverflow has literally smashed through every other Tech Q&A site so it must be doing something right.

The thing I don't like about the "reference question" idea is that initially it will probably work quite well. But then something will change or a new version will be released and then the answers on the "reference question" could potentially be incorrect.

If you base good answers on their vote count then any new answers added (to cater for the updated version etc) will probably be ignored as they'll be at the bottom of the pile.

Also, a "reference question" simply cannot cater for every problem or issue that people have. (Or should that be an individuals interpretation of their issue?) For a complex issue the "reference question" and its corresponding answers could quite quickly become an extremely long list of various answers/comments. If it looks ugly people are just not going to read it.

Also, it won't stop the stream of questions that have been asked time and time again. When posting a question you get the option of looking at previous questions that may have your solution. If someone doesn't fully read the "reference question" or it's corresponding answers for that matter, they may just think "this doesn't answer my question, my question is completely different".

When in actual fact it does answer their question they just haven't read it properly because it such a big question and answer thread.

I am not totally against this idea but I am not 100% it would be worth the effort to setup and maintain the "reference questions".

If you are saying that the "reference questions" are the best questions on SO regarding a particular subject then the existing questions are already reference questions. They are just not under a banner of "reference questions" Everyone is going to refer to them though as they have high vote counts. If people aren't referring to them - then they are not "reference questions".

Anyway just my 2p....

  • Very good points. This reflects pretty well my own doubts about the idea. The idea behind the "reference question" is in the end to distill all the best stuff from what's already there in other answers, and to collect it in one question. However, Gordon's suggestion of a digest of links to other questions would serve that goal as well. Hmmm.
    – Pekka
    Sep 7, 2010 at 21:23
  • I tend to agree with most of @Barry 's comments. If a question is somewhat old, there's a very good chance that more recently posted duplicates will get better answers if tools have changed. So: an answer upvoted 20x in the last 2 weeks may be a better answer today than the original accepted answer posted a year ago. In short, closing a question as duplicate just because there's an old answer for it isn't in our best interest all of the time.
    – Jim L
    Sep 8, 2010 at 1:49

This meta discussion misses the point a bit. In the specific case, it is not only about having reference questions, but about the type. What you are asking for is if designed reference questions are better suited for closing duplicates, than accidental reference questions.

While some Q&As naturally become reference questions because of their broad coverage, they seem to be few. And one-sided and flat reference questions are less well suited to fob somebody off when closing it as duplicate. Therefore it's certainly not wrong to have designed reference questions. In that particular case: senseful.

One more thing. There is never one answer to rule them all. People only learn from gradually aquiring knowledge. It's unreasonable to expect to turn someone into a master of a specific topic by use of a good reference answer. If they dind't use the search function prior asking, even the most excellent reference answer won't help. Doesn't mean it shouldn't be tried though.


If there is a reference question, it should be the first one that was posted (or the best one), and all other duplicates should either be merged into it, or closed and deleted. This gives all non-CW posters the opportunity to continue to gain rep from their efforts.

I understand the need and desire to eliminate duplicate content. I also believe that we have sometimes been more aggressive than we need to be at this effort, closing questions as near-duplicates of other questions, and so forth.

Certainly, when the question, "How do I parse HTML with Regex" is asked for the sixth time, some consolidation is in order. If duplication like that exists, then maybe we do need a more aggressive approach. But I haven't seen that kind of duplication, either because I don't believe it is a major problem, or because the existing system already works well at eliminating duplicates.

I find poorly worded, ill-conceived questions asked by drive-by posters a far larger problem than duplicate questions.

  • 2
    Isn't that allowing the site to become more important than the users? If there are 10000+ dups, most of them bloated - sometimes with inanity - doesn't it serve the needs of the users to have reference questions which are explicitly meant to be concise? Insisting that the first question is the one to rule 'em all places the integrity of the question base before the needs of the people who encompass the community - and SO is a community before a question base. Sep 8, 2010 at 13:39
  • 1
    ... and how do you determine when to use the first one that was posted, or when to use the "best" one? I think nominating subjects for a reference question is akin to creating a "best" one without the potential for a subjective interpretation of which version of the question is "best". It's saying "Instead of arguing over which is best and inconsistently linking to 2 dozen different questions, lets work together as a community and create the best question." Sep 8, 2010 at 13:41
  • 1
    This comment is just 'cause I like going things in threes; three is a good number. Sep 8, 2010 at 13:41
  • 1
    @Richard: I'm saying that I think it's a solution in search of a problem that doesn't really exist. The system already finds the best question by virtue of the voting system, and mechanisms are already in place to merge questions.
    – user102937
    Sep 8, 2010 at 15:06
  • Ahhh, I see. Sep 8, 2010 at 17:44

We've had this on SF for a while now for a few topics that keep reoccuring, such as users still trying to use class-based routing (or at least their terms), with a general link to CIDR for Dummies and How does IPv4 Subnetting Work? and an explanation about why we're linking them to that question, usually followed by 5 close votes.

So far it's worked well.


Carrying over my comments from that question you linked to here:

I do not think this is a good format for Reference question, because IMO a reference question will shift attention away from the actual original answers towards this one, lowering the chance for reputation gain on the original answers. Even worse, your reference question is a CW, so we don't get any reputation. Answering it offers no incentive, short of badges.

If anything, a reference question should collect the major topics and link to appropriate answers, so people still get a chance to be upvoted (like in Reference — What does this symbol mean in PHP?).

Another reason why I think your suggested type of Reference Questions isn't well suited is that it assumes it will only get quality answers. I don't see that happen. Given that anyone can answer, there is no guarantee that the answers will be top-notch stuff. An option would be to instaclose the Reference question and then merge quality answers from existing questions into it selectively. However, that requires moderator action.

  • That one makes an excellent reference and FAQ question. But I find it super unsuitable for regular nomination as -exact- duplicate. People shouldn't lazily closevote to link lists, but at least pick the actual duplicate out of there.
    – mario
    Nov 22, 2011 at 12:47
  • @mario thats perfectly possible with a Digest format. Just go there, pick the right dup and close with that instead. Although with the Operator reference its actually beneficial to close against it, because the asking OP then knows the reference exists and that might prevent him from asking about other operators.
    – Gordon
    Nov 22, 2011 at 12:58
  • 1
    Did that recently, because multiple operators were inquired about. It also has some educative benefits to do so (much like linking to the search or faq). Still feels a bit dirty under the exact duplicate banner. Nevertheless the digest format is a good reference question type. (Might benefit from CW though.)
    – mario
    Nov 22, 2011 at 13:01

Just edit the earliest version to be reasonably general and do what we're always done: close the dupes and point at the canonical version.

The suggestion that we should add new "reference versions" of questions for which there are all ready many dupes feels like saying "We're special and we get to make the right version of the question."

  • both very good and fair points. However, the kind of question we're talking about here has literally hundreds and hundreds of duplicates. I find it extremely difficult to find the reference question, because all the questions deal with specific aspects and implementations of the issue. I would be equally happy with a solution that defines one of the existing questions as "canonical" in the vote to close dialog. (But that's an entirely different discussion...)
    – Pekka
    Sep 7, 2010 at 21:22
  • 1
    @Pekka: I've learned where the early versions of the asked-over-and-over-again questions that I notice are the hard way, but that took some time. It occurs to me that the tag wikis could be very helpful here. T be sure the OP generally hasn't read it, but the closers could use it as a reference for the earliest version. Sep 7, 2010 at 21:38
  • yep, the Tag Wikis will definitely be an important tool here - with or without explicit reference questions.
    – Pekka
    Sep 8, 2010 at 9:02
  • Can you add the "tag wiki" to this answer? I want to upvote that suggestion specifically. Nov 22, 2011 at 20:59
  • 1
    The ability required for someone coming up with a CW that successfully duplicates a whole category of questions in my mind does make someone special. I wouldn't think rep should be a requirement for that. The community could be encouraged to participate in improving the answer too, which would make the community special, not a user. Nov 22, 2011 at 21:02

Personally, I like it when the system does this for me for example your exact question is on the the php FAQ

Your second question, is on the .htaccess FAQ

I think a few great approaches to solving this problem are:

  1. Improve the tag wikis
  2. Link the those canonical questions, so they appear on the top of the tag FAQ
  3. Improve the canonical questions and answers

Reference "questions" are blog posts in the guise of a question. It is very important to have references on specific issues, but it doesn't (currently) fit into SO's Q&A model. (Tag wikis are an indication that may be changing. To do it right, however, I think it would need another, similar extension to the site, at least.)

You have something you want to share, and you want to spend some effort getting it right. Why not post it to your personal blog, or submit it as an article to an online publisher such as a group blog or magazine?

SO-related blog posts work rather well. I'm sure people would be happy to get the same material linked in a comment to solve their problem. You can still close questions as duplicates, if that's what's bothering you.

  • @Gnome I have no desire to share anything here myself, I'm just the initiator of the question, so I don't really agree with your assessment. I think the question does fit rather well into Stack Overflow's Q&A scheme. The question is more about the content IMO.
    – Pekka
    Sep 8, 2010 at 15:32
  • @Pekka: Oh, then it's not a reference "question" and if it's not a duplicate, what's the problem? (You still want to share, you just want someone else to do the answering. I thought you'd be very active in evaluating each answer, too.) If it is a duplicate, why do you need the same question again? What if you gave your rep away by picking out specific questions you want to see better answered (e.g. any HTML parsing in PHP question), and offering a bounty? (In other words, why do you need a new question?)
    – Gnome
    Sep 8, 2010 at 15:38
  • @Gnome re the former, true, I would be very active there. Tthe latter is a very interesting idea. I'm a bit hesitant in messing with other people's questions, because they may need some additions in detail etc., but it might be the most fitting thing indeed!
    – Pekka
    Sep 8, 2010 at 15:59
  • @Pekka: Instead of a multi-pronged question, you can split one point into a new question if another question doesn't cover what you want and you feel it's too much of a departure to edit in. (Wouldn't be a duplicate, by definition, in that case.)
    – Gnome
    Sep 8, 2010 at 16:04
  • SO is a blog. It is also other things. It can fit just fine. Nov 22, 2011 at 21:06

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .