Are "duplicate chains" to be avoided when possible?

Question

I do a lot of patrolling for duplicates on story identification questions, and I'm looking for some thought from the group about how to handle situations where three or more questions (here labeled generically and rhetorically as A, B and C) are all judged as duplicates of each other.

In such cases, when some of the questions are closed as duplicates, one of two possibilities result:

1. a "duplicate chain," such that not all closed questions point to the same target. For example, A is closed against B (abbreviated as A->B) and B->C.

2. a "duplicate hub," such that all closures do point to the same target. For example, A->C and B->C.

Please note that THIS QUESTION IS NOT ABOUT QUESTION OR ANSWER QUALITY, which always seems to be a contentious matter due to its subjective nature. This question is about whether a duplicate chain or a duplicate hub is the preferable outcome in the abstract, and why. Take it as a given that, in the abstract examples above, C is universally regarded as the best answer. (A situation like example #1 might result if a new question C is asked and answered with high quality after the A->B closure had already been done.)

It is certainly possible to have a mix of hubs and chains in cases where there are four or more duplicate questions. My assumption is that the same principles should apply in the abstract, so please limit discussion to this simple case of A, B and C unless it is absolutely necessary to make your point.

If the consensus is that duplicate chains are less preferable than duplicate hubs, I will submit another question about preferred methods for handling chains, so please limit discussion in this question to:

• whether or not one type of result is preferable to the other
• why one type of result is preferable to the other, or why there should be no preference

Answer 1

Hub all the way

Basically all that Kevin said for the "dupe philosophy" part.

There's also a "technical" thing we have to consider here. We all have SFF accounts, meaning when we open duplicates, we're not automatically redirected when the dupe-target. That doesn't happen when you're​ logged out or don't have an account. If you open a dupe, you're shown the dupe-target (under certain conditions, like the question having no answers I think, but still). If you open a dupe of a dupe, you're shown the dupe-target of the dupe-target.

An example speaks better than words. I can't find one easily on SFF, but take So where do we go now for library recommendations? (question A) on main Meta. It's closed as dupe of Where can I ask “What library can I use to do X?” (question B) which is itself a dupe of Where can I ask a question about finding general software that meets certain requirements? (question C).

Now open question A in incognito browsing, or logged out of your main Meta account. You're redirected to question C... So ultimately, the closing of A->B is lost and kinda pointless.

So yeah, simply put, avoid dupe chains.

Answer 2

For the purposes of this discussion, I will assume that all three of A, B, and C really should be duplicates; they all ask basically the same question and all have basically the same answer. Where to draw that line has in the past been a contentious subject on this site in particular, but it's also irrelevant to the question, so I won't discuss it further.

In general, duplicates are supposed to help readers, particularly readers other than the original asker, find the best answers on the site more easily. So it logically follows that we should dupe against the best answers. A "chain" where A points to B and B points to C must violate this principle. If B's answer is better than C's answer, then C should point to B. If C's answer is better than B's answer, then A should point to C. Chains are therefore disfavored; make hubs where possible, and favor the best answer(s) regardless of chronological order.

In some cases, it's less clear which question has the "best" answers. When this happens, it's usually the case that two questions (say, B and C) both have one or more high-quality answers. The "normal" resolution for this is for a moderator to merge B and C so that there is a single question containing all of the high-quality answers. We can then choose between those answers with the normal mechanism (voting). If merging is inappropriate, for example because the questions have very different wordings, a meta discussion is probably the best way forward. Possible resolutions include rewording all of the answers under B to match C's wording (or vice-versa) and then merging, reopening one or both of B and C as separate questions (if the community decides they weren't duplicates after all), or setting up multiple duplicate targets so that A points to both B and C.

Answer 3

For story identification questions, chains are preferable to hubs.

Suppose question B has been closed as a duplicate of question A. I'm assuming they are story identification questions. Because of the nature of story identification questions, and because of the way duplicates are defined, the questions may have nothing at all in common, and the answers may have nothing in common except a story title and an author's name.

Now question C arrives, asking about the same story. Since we already have a pointer from B to A, a pointer from C to B will lead to twice as much information as a pointer from C to A.

Edit: Never Mind!

Answer 4

While it has long been a subject of contention here whether a question can be a duplicate of a different question if an answer to the different question also addresses the new question, the fact remains that many questions are closed in this manner. If A is closed as a duplicate of B because an answer to B also answers A, and B is closed as a duplicate of C because an answer to C also answers B, it is entirely possible that the answer to C does not answer A. Thus, making C a “hub” to which both B and A are closed is not a viable option, as A is not a duplicate of C in any way. If the answer to B also doesn’t answer C, and the answer to A doesn’t also answer B and C, then neither B or A can be a hub either. Therefore, there are clearly instances where having a hub simply won’t work.

Answer 5

You can't always enforce a duplicate methodology

Let's say we pick hubs as the "proper" way. What happens if someone dupe chains instead? Is there some penalty? Are mods going to correct it? Not everyone will read this Meta. And what if I'm in the CVQ and find someone doing it the "wrong way"? Do we vote Leave Open instead of closing via a question that answers the one we're voting on just because it's a chain? That sounds like a terrible idea.

If we have a dupe problem, fix the questions instead

What if we find a less desirable "chain" creating an undesirable split in dupe targets? That's the real question here, and that part is quite valid. We don't need a ton of questions floating around out there. There's a tool mods have called question merging. I mod flagged this duplicate question for a merge because this question had a really good answer not covered by the duplicate question. This produces a better outcome for everyone

• All answers can now be seen in one place
• Answers get more upvotes
• No more chains or hubs. Easier to find and vote as a dupe

You can also get a gold badge to fix "chains". Since they can edit duplicates, they can make sure everything hubs. I've done that myself on SO.