There seems to be a rapidly growing number of routine duplicate questions. For example, I collected a list of 20 questions similar to the debate about "Storing Images in Database or File System" in my answer here before I got tired of copying and pasting links. This seems to be out of control.

Would it really make sense to merge all these questions together into a single one? Firstly, the relative number of votes is partly biased by the popularity of a specific question. i.e., the oldest question has answers with vote counts like 352, 140, 99, 56, etc. If this question was merged with the rest, all the other answers to all the other duplicate questions would instantly become below the fold, regardless of their relative usefulness to the main question.

Is there a way we could begin to handle this better than fragmented exact duplicates, such as:

  • Build graphs of questions marked as exact duplicates of each other, and any graphs with more than, say, 5 nodes should get a dedicated community-maintained canonical reference question. (I wonder if such a query is already possible with the Stack Exchange Data Explorer?)

  • Build a library of these exact duplicates questions as a "Common Questions" reference tab.

  • The canonical reference question would contain a link to all related duplicate questions.

  • When a question is closed as exact duplicate of a question that is part of a "duplicate question pool", the duplicate tag displayed automatically links to the canonical reference question instead of the one the user closing the question selected.

  • When displaying possible duplicates list as a user is entering a question, prefer questions from this "Common Questions" list as they are more likely to be related and catch the user's attention.

  • 2
    I don't have an answer to your question, but the "answer" you linked to is not in any way a real answer. It contains absolutely no informational content. If anything it should be a long comment or several comments. The point is that you should never post an answer containing only links, especially if they are all links within the site. – murgatroid99 Jan 26 '13 at 0:09
  • There is a tendency for a question to be closed as a duplicate if it shares an answer with a significantly different question. Another common occurrence is for a question to be closed as a duplicate if there is another very similar question, even if the answer of one does not answer the other. A question should be closed as a duplicate only if (a) the questions are essentially the same, especially in the sense of a search, AND (b) the answer of one applies to the other. Merging questions as mentioned here may be a good idea, but I suspect many non-duplicate questions will be deleted. – xpda Jan 26 '13 at 0:42

There are a lot of good ideas here, but this one seems backwards to me.

The canonical reference question would contain a link to all related duplicate questions.

The canonical reference question should fully explain the problem in detail, and it should contain a complete answer. There really shouldn't be any need to link away from that page to others. All of the information should be there. If it isn't, it either shouldn't be the canonical Q/A pair, or the information should be added.

I like the idea of building graphs (by automated means, if at all possible). Theoretically the graphs should not have any paths longer than one edge back to the canonical question, but we know that's not the case now. Question A gets closed as a duplicate of B, which later gets closed as a duplicate of C, for example. It would be nice to have A link directly to C (or at least to both B and C) when these paths are detected.

Build a library of these exact duplicates questions as a "Common Questions" reference tab.

This sort of information is sometimes collected on tag wikis. Unfortunately, we haven't figured out a way to get people to use them. (I'd love to see more collections of links like the tutorial section on the Scala tag wiki.)

  • 2
    The system must already make parts of such graphs internally (though I don't know if it stores them, or just calculates them anew each time they're needed). After all, it prevents cycles of duplicates that loop back on themselves. Therefore, features relating to this might be relatively easy to implement. I agree that if X is duped to Y which is duped to Z, then the possible duplicate banner on X should list Z. I think it should list Y as well, though; sometimes you need Y to understand how X is a duplicate of Z, or how to use information in Z. – Eliah Kagan Jan 26 '13 at 1:44
  • I agree that it isn't feasible to link back to all related dupes, though maybe eventually they can have a sub-page of the tag page where those links are listed categorized by canonical headings. – Lance Roberts Jan 26 '13 at 6:20

You must log in to answer this question.

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