11

So far, we haven’t closed many questions as duplicates (more precisely, we don’t have many questions closed as duplicates which haven’t been deleted), and we haven’t decided how we want to handle questions in detail.

Specifically, how should we define duplicate questions?

  • Questions which are exact duplicates of another question.

or

  • Questions which are answered (fully) by answers to another question.

or perhaps something else?

Stack Exchange in general seems to use the latter definition; this has resulted in the text used in “duplicate” boxes:

“Duplicate” box

This question already has answers here

followed by (a) link(s) to the target question(s).

This came up with What did the 8086 (and 8088) do upon encountering an illegal instruction?, which seemed to me to be answered by my answer and supercat’s to another question, Use of undocumented opcodes — and the top-voted answer to the first question refers to the latter question.

After further reflection I no longer think the two questions are duplicates of each other, or even that the answers to either are complete answers to the other. However the question remains, and I think it would be better to decide it now rather than later (obviously nothing is set in stone and as the community evolves its practices will too).

1
  • There are 16 undeleted, and 8 deleted, closed-as-duplicate questions at time of writing.
    – wizzwizz4 Mod
    Jun 19, 2019 at 13:59

4 Answers 4

Reset to default
10

A question isn't a duplicate just because it is answered by the answer to another question. It's only a duplicate when it asks substantially the same thing, to the extent where any good answer to either question must necessarily be a good answer to the other.

3

(Not a full answer or a guideline, just a starting point from the two definition Stephen did)

I do not belive that this question can be answered at all - at least not in a serious manner. It's the classic "Everyone will know that it's a duplicate, when he sees it" case. Like what is a bird ... seames easy, until you note that a Kiwi is as well a bird :))

Questions which are exact duplicates of another question.

Not sure if it will be ever be the easy case of a verbatim copy. In reality a question needs to be evaluated what is asked and then compared, as there are many ways to ask the same question. Just think about some hardware issue around some 74xx in an Apple - the very same can come up in an IBM-PC setup. Even more so in software or any other area.

Questions which are answered (fully) by answers to another question.

Even less. Keep in mind, people often (*1) write lengthy answers with additional/background information to show a wider aspect the question is related. Such an answer may as well answer some other (later asked) question in full. Nonetheless the (newer) question may stand in its own right, even thru it has been answered elsewere.

So no, an answer already given on another question will not invalidate the new one - if it is different ofc.

A helpful solution here would be writing (*2) a specific answer for the new question, but also give a link to the related question/answer and pointing out why it's related.

*1 - Yes, guilty as charged.

*2 - This may as well be just copying and reformatting the relevant passages to target the information requested by the second (new) question.

3
  • There can be many ways to phrase a Q. But: If copying a partial answer suffices the it is a dupe. If the newer Q is formulated very differently, then it is probably an upvote-worthy dupe, very much so. If searching for the terms employed yields the partial A here (lack of difference), then a downvote might be in order for lack of research effort? // The way this system is currently designed the link to related Q&A should be paramount, but alas is not (as it's often just in volatile comments). The goal is to have a person have a Q (even from outside SE), come here, and find the answer.
    – LаngLаngС
    Jun 19, 2019 at 16:11
  • 1
    @LangLangC Not necessary. Questions often get answers that contain related information beyond the core question, and it can as well be this additional information which would be a sufficient answer for another question - thus one answer (or parts thereof) can quite good fit multiple questions - just this doesn't make the question a dupe - it makes tha answer versatile.
    – Raffzahn
    Jun 19, 2019 at 18:28
  • Note the dupe boiler plate: "This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question." (or re-phrase so that it's clearly no longer a dupe?) –– Without concrete examples to compare, I'd maintain that the general case you outlined would 1. necessitate the linking (achieved via dupe) 2. it's the As that count, if it's "versatile", good, if it needs rewrite -> no dupe 3. As can be also 'too versatile' (lack focus, not often seen with upvotes; or here?)
    – LаngLаngС
    Jun 19, 2019 at 19:38
3

You are right, we don't have many questions that have been closed as duplicates. The simplistic answer is, of course, that not many have been flagged as such.

As to what constitutes a duplicate, several StackExchange sites have agonised over this issue. For example:
SF&F, EL&U and SO to name but three.

And of course on the main meta, several questions all marked as duplicates of...
Changes to "close as duplicate" (part deux)

In general, it is not easy to spot duplicates. Nobody can be fully cognisant of all the thousand-odd questions that could be construed as a dupe-target. People tend only to flag duplicates if they remember a particular earlier question - that is certainly how I flag duplicates.

It is possible to ask questions in different ways. So, two people seeking the same answer may ask very similar looking questions. This doesn't necessarily make them duplicate questions. Later people seeking the same information may find just one of the questions because of how they search and our site has been helpful to them.

I don't believe our site is suffering because of our number of duplicates. I don't believe that we need to make a conscious effort to change our ways in this regard. Without doing any sort of data trawl, I would suspect that all new SE sites start out with few duplicates and get them at increasing frequency as the number of older questions grows.

As to the questions that you mention, I agree they are not duplicates and that the answers are relevant to both. In this, and any other similar situation, putting comments on each question with links to the other would be constructive.

3
  • 2
    Minor nitpick: regarding "Later people seeking the same information may find just one of the questions because of how they search and our site has been helpful to them.", this is the whole reason we close as duplicate, instead of deleting, such questions.
    – wizzwizz4 Mod
    Jun 19, 2019 at 13:57
  • The 'putting comments' reads as explicitly violating policy. If it is a dupe it needs dupe-close. If it is linked, it needs stable links, ie in the post itself. There are good dupes and bad dupes. Those are good that ask the essentially same Q quite differently, adding a path or sign post towards the A. And bad dupes, just displaying lack of prior research/laziness. One should get an UV the other zero or a DV? (Disclosure: I do not know/understand how the system would allow "mergers": leaving just the dupe Q and migrating existing As to the 'goal' post?)
    – LаngLаngС
    Jun 19, 2019 at 19:44
  • @LangLangC Yeah, that's the gist of it.
    – wizzwizz4 Mod
    Jun 20, 2019 at 8:21
1

I think that both those options of looking at duplicates have flaws:

Questions which are exact duplicates of another question.

No question is likely to be an exact duplicate, there has to be some judgment as to the similarity.

Questions which are answered (fully) by answers to another question.

This could lead to the scenario where a question is precluded just because it's answered in a totally unrelated question.

By way of example (since I live mostly on SO): let's say there is a question Does Java have undefined behaviour? and one of the low-voted answers states that an example of UB in C is what happens when a signed variable wraps around.

If someone then found that their variables in C became strange values when they got too big, and wanted to find out why, there is no way that they would think that a question about UB in Java is going to help them out (especially if the top-voted answer is likely to be just "No", without mentioning C at all).

The guideline I tend to use for dupe-close is if either:

  • the potential dupe seeks a proper subset of the information sought by another question, or
  • if there is substantial crossover in the information sought and it is all provided by answers to the other question.

Both of these guidelines relate to the question primarily and I believe that's important since you're closing the question.

This meta-question states the box added to a question after dupe-close contains the test "This question already has answers here", in support of the "Questions which are answered (fully) by answers to another question" approach.

However, the actual description that appears when you attempt to dupe-close is (my emphasis):

This question has been asked before and already has an answer.

Note that my guideline covers things like (returning to the RC subject matter) What do I need to watch out for when turning on a ZX-80 that has been unpowered for twenty years?.

I would consider that a dupe of a question asking substantially the same thing about a C64 or TRS-80. It's a different machine so technically not a proper subset of information being sought.

But it does fall under the substantial crossover clause and answers will almost certainly be the same ("watch out for capacitors").

Anyway, that just my two cents worth and, hopefully, you won't feel like I've overcharged you :-)

You must log in to answer this question.

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