I know we've already had a question about duplicate posts on meta about it (the irony! :) but that question doesn't address the issue in exactly the same way I am wanting to address it:

One of our users is being very active in tagging what he feels to be duplicate questions. While I'm glad to have such a passionate user of the site, I am now starting to dread moderating as it seems every time I go to moderate I see his tagged-as-duplicates forcing me to review each one and make a decision on them.

The reason I dread it is not for the work but because I have very mixed feelings about closing duplicate questions. The original question may be very similar, but the exact use-cases are often different and the users are clearly different people. Further, older questions do not get the attention that new questions get, so often the only way to get a question answered is to ask a newer, identical one. So closing "duplicate" questions seems very problematic to me. And when a question is closed as a duplicate it feels very much like a negative to the original poster, and why embrace a mechanism that makes (mostly) new users feel bad about their first experience? (Talk about about the number of frigid users this will produce! But I digress... ;)

Here's what I propose: When a user finds a "duplicate" post they make yet another post, mark their post as community wiki and they do their best to write a perfect SEO-friendly title, a perfect summary of the question, tag it as <canonical> and add their best answer culled from the answers to the other questions.

In addition, at the very bottom of the question they would add a bulleted list of the other questions exactly like the following:

This question is the canonical version of these questions:

- [Question Title 1](http://wordpress.stackexchange.com/questions/123/)
- [Question Title 2](http://wordpress.stackexchange.com/questions/456/)
- [Question Title 3](http://wordpress.stackexchange.com/questions/789/)

Then later, StackExchange can add functionality to enable a checkbox for <canonical> that would allow the poster to enter the URLs for the related questions. This could also have the result of the following within the StackExchange mechanism:

  • Setting the canonical header in the HTML of the other questions to point to the canonical question,
  • Merging all search results that match the related questions, and linking the merged search result to the canonical question,
  • Listing the links to the related questions at the bottom of the canonical question,
  • Linking to the canonical question on the page of the related question but above the text for the related question, and
  • Setting a flag for moderator review to ensure the the other questions are indeed related.

This change could:

  • Remove the "problem" of duplicate questions, turning them into a positive,
  • Institute a formal structure for canonical questions, and
  • Encourage and empower users to curate a canonical reference.

There really is nothing that's part of StackExchange network that does a good job of encouraging users to curate a canonical reference yet finding ways to create a canonical reference for each common question should IMO be the goal of a Q&A site. Giving users points for the canonical question and answers in some form could also really supercharge curation and result in a way to use the weeds to cultivate beautiful flowers in the garden (where weeds=duplicate questions, flowers=canonical questions, and the garden is the StackExchange mechanism and network of sites.)

Thoughts?

share
    
I'll stop flagging dupes! –  anu May 16 '11 at 10:26
1  
@anu - LOL! I wasn't trying to get you to stop. :) I was looking for a better way for the StackExchange mechanism to handle duplicates. Please don't let me discourage you from being active. –  MikeSchinkel May 16 '11 at 17:55
add comment

3 Answers 3

I agree that we need more canonical answers, and I am already thinking of some areas where I could provide some. One problem is that they might turn out to be complete manuals, but a manual where there was none before ("missing" parts from the Codex) is better than nothing at all. I don't think they need to be community wiki, since everyone can edit or at least suggest an edit for posts now, and I see no harm in giving rep for a good canonical question and answer.

But it is unclear to me why the links from the canonical question to the specific question can not be made the other way around, from the specific question to the canonical question. This system already exists: mark it as a duplicate question. It adds a link, and maybe it merges search results, but if it doesn't it's an easy extension.

It is unclear to me what would happen if there already is a canonical question, and a new specific question pops up. Should I close it as duplicate of the canonical question? Because that is what we are trying to say here: "I think you can solve this question by reading the answer to this question." Implicitly, we are also trying to say "If you can't solve it after reading that question, please edit your question to clarify what you tried and what did not work." Just "Possible duplicate of" could indeed be too short to convey this message - but it's a comment, and you can add an extra one if you want to. I had this experience with a recent question: anu and I misread it and thought it was a simple duplicate question, but after a discussion with the author the question was edited to become more specific to an issue not already discussed in the other question.

To summarize: I think we need more canonical questions, but I don't think we need an extra system to replace the "duplicate question" voting system we currently have. However, I understand you feel less comfortable with it because you have a "binding" moderator vote, whereas when I vote it needs four other users to agree with me. The solution there is to upvote anu up to 3000 rep so he doesn't need to flag a post when he thinks it doesn't belong here :-)

share
    
"This system already exists: mark it as a duplicate question." Two things: (1) does marking as duplicate not also hide it? I'm not sure. (2) Posts marked "duplicate" are currently viewed negatively by the community; that would need to change and the community culture won't change except for overt action by StackExchange. "Should I close it as duplicate of the canonical question?" Not close it, associate it. "The solution there is to upvote anu up to 3000 rep..." But then @anu will do what I am trying to ensure does not happen. :-) –  MikeSchinkel May 16 '11 at 17:49
    
" I think we need more canonical questions, but I don't think we need an extra system to replace the "duplicate question" voting system we currently have." Isn't the latter what I'm proposing? –  MikeSchinkel May 16 '11 at 17:50
    
@Mike: In my view your proposal consists of two parts: 1) we need more canonical questions and 2) we need to link specific questions to canonical questions. My idea is that we already have a way to do 2), and that is marking it as a duplicate question. I don't see enough difference between marking something as duplicate of another question and associating something with another question: in both cases, the original question is closed, no? –  Jan Fabry May 17 '11 at 7:03
    
The duplicate system has the advantage that it is a community action, with five votes needed to close it. Because we don't have enough high-rep users this currently requires a lot of moderator intervention, which is not ideal, because more viewpoints are always better, but this will diminish over time. To deal with the negative stigma of the duplicate close reason, would it be an improvement if the "Possible duplicate of" sentence was changed to something longer and more explicit like I proposed? –  Jan Fabry May 17 '11 at 7:08
    
"In my view your proposal consists of two parts:" Yes, you got two (2) of the three (3) reasons for my proposal. The 3rd is my desire to see duplicate questions change from a negative to a positive and allow people to ask a question that is similar to another in order to be able to get their own specific questions answered and to keep from having their question closed and thus avoid negative feelings they will get. To me #3 is most important. –  MikeSchinkel May 17 '11 at 18:10
add comment

The original question may be very similar, but the exact use-cases are often different and the users are clearly different people

Agreed, you want some duplication but not too much. See

http://blog.stackoverflow.com/2010/11/dr-strangedupe-or-how-i-learned-to-stop-worrying-and-love-duplication/

In addition, at the very bottom of the question they would add a bulleted list of the other questions exactly like the following:

Why would anyone need to do that, when Linked Questions in the sidebar already does this completely automatically?

http://blog.stackoverflow.com/2010/04/new-linked-posts/

All that really needs to be done is decide which question is canonical, make it great, then as new duplicates come in, close them as duplicates of that. (You could also merge the questions if you like, but it shouldn't be required so long as you catch the duplicates early.)

There's no getting around the "requires human judgment" part of this, but all the tools to do it already exist.

http://blog.stackoverflow.com/2011/01/the-wikipedia-of-long-tail-programming-questions/

share
    
"Why would anyone need to do that, when Linked Questions in the sidebar already does this completely automatically?" It's just not as obvious that they would be the related to the canonical ones. "close them as duplicates of that." That's the part I was proposing to avoid. –  MikeSchinkel May 19 '11 at 16:33
add comment

I want to revisit this question, to see if we can perhaps come up with a list of questions for which it would be beneficial to provide "canonical" answers.

My list would start with at least these two:

  1. Custom loop/query pagination

    I probably average one question per day that I close-vote as a duplicate of "why doesn't my custom loop/query pagination work?" variants. We have one question/answer that has gotten the most votes, but it could probably use some work to ensure that it covers all the bases, to serve as the "canonical" answer.

  2. Unexpected output ("headers already sent")

    For the second, the issue is almost always that the asker used a WordPress function that echoes rather than returns its output, causing the unexpected behavior. The biggest benefit of having a canonical question/answer (perhaps with a list of return/echo WordPress function pairs, and relevant Codex links), would be to avoid having to answer the same thing for the dozens of potential WordPress functions that can cause the problem. It is a legitimate question, but quite honestly, it gets boring to answer the same thing - especially when the answer is no farther away than a given function's Codex entry (something else we want to try to avoid).

  3. jQuery no-conflict & proper script enqueueing

  4. Shortcode + echo

I am curious to know what other topics/questions the WPSE community thinks would be beneficial as "canonical"? I'll be happy to help contribute content to the canonical answers.

share
    
Shortcode + echo and $ doesn’t work in jQuery are quite popular too. –  toscho Dec 4 '12 at 9:33
    
Good ones. Added to the list. –  Chip Bennett Dec 4 '12 at 12:28
add comment

You must log in to answer this question.

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