meta discussion: should we close resolved discussions?

[lucascolley]

thanks to @matthewfeickert for bringing this to our attention! https://discord.com/channels/1082332781146800168/1399805113799020627

Preface

Getting some duplicate questions from users is inevitable, but getting less of them is ideal. We can roughly categorise what a user has done before opening a discussion/issue as so:

1. No searching at all, ignoring any guidance in issue/discussion templates. Here, duplicate questions are inevitable.
2. Some searching, likely including (at least an attempt to) follow guidance in issue/discussion templates.
3. Thorough searching, covering closed issues & discussions, discord, maybe even other forums.

The success of cases of (2) in preventing duplicate questions is influenced by our repo setup (not so much for cases of (1) or (3)). We should try to set it up in a way that helps prevent duplicates.

Closed discussions

One potential problem with closed discussions is that GitHub applies the is:open filter by default on the discussions tab. People have raised this to GitHub as a potentially problematic default which could harm discoverability of answers (https://github.com/orgs/community/discussions/3097#discussioncomment-5392051), but I'm not aware of any plans to change it.

Just leave all discussions open?

Leaving all discussions open would immediately solve this problem of discoverability under the default filters. Of course, we could still close irrelevant discussions that are unlikely to help with any potential future questions.

Why might we not want to do that?

The flip side of the coin is that leaving resolved discussions visible with the default filters harms visibility of unresolved discussions under those filters. Visibility of unresolved discussions matters from a maintainer side, so that important bugs/feature requests don't get lost and ignored.

A potentially good argument against this point is that we could create a new label (to mark either resolved or unresolved discussions). Then we just need to apply a filter to get a focused list of unresolved discussions as a maintainer (which isn't a big ask). There would still be some manual labour in applying these labels, but probably not much more than ensuring discussions are appropriately open/closed.

Personally, I much prefer using the intended feature from GitHub to track resolved things, than making up our own. I feel like introducing more labels and open discussions on the default page will increase cognitive load, while closing discussions decreases it. However, that preference might not be pragmatic for the reasons outlined above.

Is there a good alternative?

One idea I had was potentially updating our issue/discussion templates to ensure that they clearly direct users to search for potential duplicates to a reasonable level of rigour. I think it should be possible to do this whilst keeping the ask of users low, by e.g. including URLs for appropriately filtered searches in the templates.

I speculate that this would largely solve the issue, while still allowing us to use the resolved feature as intended by GitHub. I don't think that there would be a large number of users who would search discussions under the default feature, but would fail to follow a clearly signposted and low-effort instruction.

I propose that we try that approach and continue to make use of the resolved feature as intended, then make an empirical judgement about my speculation that it would work pretty well. But I thought we should be transparent about that in the first instance and open up to other opinions!

---

cc @ruben-arts for visibility