Automate closing of stale topics

Since necro-posting has been such a recurring issue, would it be possible to automatically close an Ask Fedora topic where a solution has been marked and there haven’t been any new replies in a two week time, with an automated message encouraging a new topic be created if someone has a related problem?

Not that we need more motivation, but this would also help us focus all the more on getting a marked solution for topics as well.

4 Likes

I think if there’s a Solution, then a month at most. Also, how Discourse displays the message that a solution has already been found could be more emphasized. There have been a ton of Necro’s recently and that does in turn become a lot of Mod work.

If there’s no solution, I still think topics should be closed after a shorter than 9 month period. Fedora moves too quickly for that to be the period of time topics are opened.

1 Like

I tend to agree with this.

I agree, even having a relatively rapid time like 24hrs after solution marked would be fine IMO.

1 Like

Thanks for bringing this up. I was thinking the same thing. At one or more of the old Ask Fedora sites, there was auto-closure for “Solved” posts, but I don’t think that’s really the right trigger. Sometimes, those should stay around even though someone hit the checkbox — the solution is maybe not the best one, or new information may come out. And, to me at least, “closed” signals “no longer relevant” more than “this will help you!”.

But anyway,there are several possible approaches I can think of offhand:

A. Use the category auto-close feature:

A screenshot of a webpage with a form that includes a dropdown menu with the options "18" and "months" selected, and a message below the form stating "Don't close until the last post in the topic is at least this old."  (Captioned by AI)

B. Slightly different auto-close feature from the “solved” plugin

Option for: Auto close topic (n) hours after the last reply once the topic has been marked as solved.

As noted, I’m not a fan, but I could be convinced.

C. Do a manual clean-up after each EOL, or annually, or something

In this case, moderators could bulk-select and close posts.

D. Use Discourse Automation scripts

Discourse Automation is a plugin which can run behind-the-scenes scripts on various triggers. (This is what’s adding the polls in Change Proposals.)

Unfortunately, there’s not a way as a hosted customer to add or modify the available scripts,[1], and the existing Close Posts script currently only has triggers for closing a specific topic at a given time or for closing “stalled wikis”.

So that would need to be extended.

E. Someone write an external script that uses the API to automate something. (Like @issuebot.)

I do not volunteer. :classic_smiley:


  1. although I have never tried writing one and asking for it to be added ↩︎

1 Like

Also, in addition to closing posts, there are some other states that we could consider.

Closed, or Archived?

See The difference between Closed, Unlisted and Archived topics - moderators - Discourse Meta for details, but in short, archived topics are frozen — can’t be deleted by the poster (mods still can), can’t get likes, and won’t be included in the “top” view (which I’ve renamed to Active because “Top” is confusing — see Ask Fedora - Fedora Discussion).

I think Closed is right, but I wanted to bring this up.

Closed, or Moved?

Take a look what we currently do for Common Issues — old ones get (manually) moved to Archived Common Issues. (Manually because I was going to make @issuebot do it, but haven’t gotten around to it.)

The advantage here is that categories can have different search priorities. For example, Common Issues is set to “Very High”, while Archived Common Issues is set to “Very Low”. (And Ask Fedora is "High.) This means that if searchers don’t go out of their way to specify the archived category, old things are less likely to come up.

It also avoids my (possibly weird and wrong?) instinctive reaction that Closed is inherently negative signal.

I don’t currently remember how “Closed” influences searching. Last year (this was only last year?) I suggested upstream that closed:solved should get a priority bump, but I think that’s not quite what ended up being implemented.

Sorry, you will never have the time to read all the answers within 24 hours. If yes pls. tell me how :slight_smile:

This would be very easy for solution hunters that they just create topics solve them and let them die after 24 Hours. This would also give much more work to manage this kind of problem.

I agree that 24 hours is too short. 2 weeks to a month would be the sweet spot, IMO. A month almost seems too long, though.

While this could work for most cases, there are plenty of legit topics that go on for months, especially when debugging something that is kernel/hardware related, for example.

This seems like the best immediate option, IMO, so long as hours can be set to at least 336 and be specific to Ask Fedora.

This is sort of what we do now, more or less. More reactive without intentionality. Doing it in bulk makes it more likely we might get some things wrong and people might complain they were treated unfairly and we won’t be able to blame the automated script.

This is the best longterm option as we help Discourse and everyone else using it, but do any of us have much experience in contributing to Discourse or the bandwidth to do it? I do not.

Normally, this stuff is right up my alley. A CronJob task in OpenShift somewhere that runs a bash or python job in a container - sure. I have to be honest that I’m not in a place at this very moment to take that on, but if we don’t have another agreeable option in a few more weeks, nudge me and I might be willing then.

1 Like

I tend to say align it with one release cycle: 6 months (I tend to say there is no need to invest efforts to align it with release dates and release tags - just let it be 6 months at the time of opening the topic). Sometimes they can be useful to avoid duplications and to merge data, not often, but the moderation efforts are low for a few months more. Within 6 months, it remains a realistic possibility that an unsolved topic’s issue may reoccur at someone else with the same origin and that even the user who opened it may profit from it being solved at later time (and data from multiple users can be gold in complex problems!).

It is more the very old ones where the release number and age of the topic make clear that it makes no sense to let it further develop. 6 months already includes a strong mitigation for the “proven obsoleted release” topics. I’m a fan of the least invasive means, and the moderation efforts are limited with allowing 6 months :slight_smile:

I am 100% for your proposal, but we have been surviving without it as well: so I tend to give more time rather than less. So I think 1 month should be already a strong mitigation compared to the current condition.

Also, take into account that sometimes (although seldomly) users become aware that they accidentally only bypassed the issue but not really solved it (so that it then re-appears in other manifests or so).

1 Like

I’m OK with one month. Yes, we’ve survived, but there’s always room for improvement and fixing this would allow us to spend more time/energy on actually helping people than reacting to necro posts. Individually, they’re a minor nuisance, but the annoyance builds up over time. It’s an honest mistake since many forums operate under different rules. An automation can help us better steer behavior in the way we would like and besides, it doesn’t seem like the best experience for someone who necro-posts as one of their first interactions to be advised that they shouldn’t have engaged in that way and to open a new topic. This is an opportunity to improve the experience for both moderators and regular users.

Absolutely :wink: My point was in this respect more to focus on the least invasive means: if we are unsure if 2 weeks or 1 month before auto closing, then make 1 month.

There might be cases where it can even save time to not have closed a topic. My thought was that within 1 month, this can be realistic for topics that are (likely to be fully) solved.

100%

1 Like

Note that the auto-close feature can be set to reset when there’s a new reply, so the countdown only starts when the thread has really died down. So that should handle that.

I need to do some SQL to check but I think there’s probably a lot of not-solved stale questions – either never marked solved, or hard to resolve, or person went away.

But that said, I tried 876,582 and that apparently works, so no problem on that duration requirement.

1 Like

My overall thoughts are that a thread that has a marked solution should be closed in about 30 to 60 days, no longer.

For threads that have never had a solution marked then 6 months max because of the release cycle of fedora makes it almost certain that the version in use when the thread was started has seen a newer release and at 12 months it is 100% certain that the version is now EOL (except for topics on beta versions of course – but 6 months is still way far beyond whatever software version was in the topic.)

The recent spate of necro posts has seen threads reopened for topics that were dead 2 or more years back.

3 Likes

Many topics do not get marked with solutions.

Do we need to pick a solution when one does exist but OP does not set one?

Personally, I tend to ask the OP to pick one. If they don’t, and if I come to the conclusion that this topic can help a broader audience, I then select one. But I do that only under the condition that there is the one distinguished/clear solution that was confirmed by the OP. I also add the sentence that the OP is free to change the selection if they feel another to be better.

3 Likes

I have asked OP’s to pick the solution and most are very open and eager to do so. I have closed topics that have had solutions marked but meandered into a number of divergent comments after, or were necro’d two years later. I’m not certain (socially speaking) if that is the nicest thing to do but, at times around here we get chatty. Basically, I think it is sort of expected of the moderators to nudge the community on occasion to stay within the expected scope. Nvidia cards come to mind as one of those every cycle repeat issues, god I wish Nvidia would take the Linux community seriously enough to at least make drivers better for their hardware in the kernel.

3 Likes

I think it’s fine to close questions that have wanderered off topic. And, if it makes sense, direct people to The Water Cooler — or even move posts to existing or new topics there.

4 Likes

I suspect many necro-posts result from users finding a solution an not following up with information that would help others with the same issue, including users “solving” a problem by switching distros. This is a community involvement issue. It should be made clear that switching distros can be the appropriate solution for some users and encourage “Solved by switching distro” posts.

“Solved” often fails to adequately describe the situation, for example, a temporary workaround for a problem that requires a proper fix from upstream . In that case, it would be useful to keep the topic until a proper solution appears, with a note that it is waiting for upstream fixes.

Having several options in addition to “Solved” – maybe “Good Fedora Solution”, “Workaround that needs followup”, “Solved by switching distro”, and “Really Ugly Hack”.

1 Like

Interesting idea — but currently, the Solved plugin just has the one.

1 Like