A discussion about closed (and potentially useful) posts on Stack Overflow

There is a feature for "Exploratory or open-ended" posts, but Stack Overflow ignored it very shortly after implementing it, left it in a terrible feature lacking state, and then very recently fooled hundreds of users into posting comments in there, and continue to do so... Why not focus on fixing that issue if you are concerned about off-topic content on the site? That area of the site could have been an excellent area, but now it's an utter mess...

That portion of the site reminds me of the current state of Birmingham. Perhaps you could clean it up and get some much needed functionality and guidance in there?

The first thing I want to highlight - because I'm not sure from the OP text that it's been properly understood - is that being "on topic" is not sufficient for a question to pass muster.

Likewise, if you’re a curator, voting-to-close or flagging content, you’ve certainly come across people posting “off-topic” questions, and maybe even wrestled with gray-area questions that seem valuable yet don’t quite belong.

No, not really. Figuring out the appropriate breadth of scope for a question (especially for common, introductory-level topics) is far more stressful. It's also sometimes difficult to figure out whether the OP just wrote some non-working code, made a simple error in logical reasoning, and failed to attempt proper debugging; or whether there's an actual underlying conceptual issue that prevented that debugging effort.

Is the “off-topic” content list posted above complete? Does it match what you flag or see closed as off-topic? What examples are missing? Are any not what you think of as “off-topic”?

It's simply wrong. The problem with exploratory questions generally is that they need more focus, not anything to do with topicality.

Specifically regarding comparison questions: these could be fine if the OP provides a single, objective metric for comparison, but then wonders why OP doesn't just try it and see which option performs better on that metric. But when people ask about alternatives like this, I think what they're usually really trying to figure out is a process for choosing between the alternatives - a heuristic for making the decision based on personal needs, without baking the assumption of those needs into the question. And, with appropriate care, I think this sort of question can already perfectly well fit within policy. The problems occur because people want someone else to decide for them what factors are most important, or how fast / memory-efficient etc. will be sufficient in a given context, etc.

Broadly, and notwithstanding flaggable content like spam, we can classify the reasons we close questions in three categories:

1. There isn't a new question that actually benefits from being asked:

   • Duplicates
   • Not reproducible or caused by a typo

2. There's a problem with how the question is asked, or with how much is asked:

   • Needs details or clarity
   • Needs more focus
   • Needs debugging details
   • Not written in English

3. There's a problem with what is being asked (off topic):

   • Opinion-based
   • Not about programming
   • Seeking recommendations
   • Belongs on another network site

We already have these definitions and we don't need to rehash that.

What characteristics make it incompatible with the site’s goals?

I think this follows pretty naturally from what the goals are. We care about topicality because a) putting things in the right place makes them easier to find; b) the "fixing broken windows" effect.

Which examples spark debate, or are tricky to call off-topic or on-topic?

Historically, there's been quite a bit of argument over where the line is between "shell scripting" (i.e. actually writing a program using Bash, Powershell etc.) and just operating the computer at a terminal.

do you feel the boundaries of what’s considered on or off-topic have changed, perhaps gotten stricter or more lenient?

No; what changes is the amount of resources available for enforcement vs. the volume of new questions, and the importance placed on that enforcement. The actual rules being enforced have changed quite little over the last 10 years or so. But now, instead of just grumbling about the constant torrent of low-quality questions, there's a greater understanding of what we're trying to accomplish (have always been trying to accomplish!), what the policy is, and how the policy aims towards the goals.

Have perceptions or rules around closing questions for shifted over time? Looking back on your experience with Stack Overflow, do you feel the boundaries of what’s allowed have changed, perhaps gotten stricter or more lenient? What might have driven those shifts (e.g. community needs, site goals, moderation tools)? Share your thoughts or an example that stands out.

Beating the dead horse yet again: yes there used to be a stricter requirement that the poster had a minimum of knowledge about the technology they are asking about. Originally SO was a site for enthusiast or professional programmers. A programmer being a person who knows at least the basics of at least one programming language.

There has been a shift to allow unregistered users at day 1 of their programming studies to post their "hello world" questions, in order to maximize site traffic. If you tell them "the answer is actually mentioned on page 1 of your programming book" you are being rude and should be slapped in the face with a welcome wagon. Whereas it is apparently not rude to repeatedly ask unpaid volunteers inane questions that anyone, programmer or not, can answer with a minimum of research. The definition of rude has definitely shifted from "impolite, disrespectful, judgmental" etc to "person who tries to steal our site traffic".

Overall I feel that the community has always been pushing for quality and the company has always been pushing for quantity. One blatant (dead horse) example is when we tried to Delete the list of random books? This was repeatedly closed and deleted by moderators and even SO staff. And then from somewhere within the company there suddenly came a scream: "Noooo they are trying to delete the click bait post with a million views!" And there you go - post undeleted and locked. It is still sitting there, serving all our philosophical and life advise needs... All that is missing is a post about which cute cat pictures that are most important to programmers and maybe also one about programmer porn, to maximize the amount of clicks.

This quantity over quality policy has finally backfired severely due to chat bots being to answer all them "hello world" questions - this is actually something AI is good at. And so panic and layoffs within the company, as the day 1 programmer students consisting of some 80% of the site traffic left. Quantity turned out to be a sand castle.

But perhaps (a completely overhauled) Discussions + chat could lure those kind of users back. Having one single site to cover every type of programming concern (within reason) has its appeal. Lessons learned from Q&A is that quantity over quality isn't a sustainable business model in the long run and so Discussions and chat need to be held to relatively high quality standards too. Discussions is nowhere near quality, it is not even near a MVP. "It is popular so it must be good" No! Haven't you learnt anything? Cute cat pictures are also popular. For now, anyway, maybe it's dog pictures tomorrow.

As mentioned many times before, the Q&A format here on meta is entirely unsuitable for discussions. 50 guys having monologues in parallel where detailed responses in comments are discouraged is not very productive. So why not have it as a goal to migrate the whole discussion part of meta to Discussions? In that case, when such a migration is at all possible, then we can probably say that Discussions have reached MVP.

The examples you have chosen do not represent adequately what we consider off-topic questions. Questions are usually considered off-topic when they are not about programming, e.g. what should I feed my cat or why isn't my WiFi connecting.

Homework questions are generally on-topic but poorly received and may be closed for lack of clarity or focus.

API questions and theoretical questions are fine if they are to do with programming.

There are certain grey areas such as CLI/shell questions, docker, kubernates, apache. This isn't the worst place to ask such a question but technically most of them have nothing to do with programming.

There are also questions which seem like they should be on-topic but aren't, e.g. azure pricing or Google store support. We just can't provide a definite answer to them. These are truly unfit.

Then there are questions that are either typos or are about a code error without the steps/details necessary to reproduce the error. They are off-topic because answering them is unlikely to help anyone else.

Many historical locks have been applied to questions about Linux commands. While the topic is not programming related, it is used so often by programmers that keeping them here is useful. We'd rather that they are asked on more appropriate sites.

I want to focus on an area I feel has created a grey area for some; Collectives.

Collectives designed to group certain tags, so that experts in that technology, and users using it, can find a range of content in it. It has is own leader boards, articles, and can have experts and recommended answers defined. I don't think any of these are bad ideas, and if managed correctly, do add a small benefit.

What is confusing is the focus of some of those Collectives; AWS, Google Cloud, and Azure immediately spring to mind here. When I think of these technologies I don't think Programming and Software Development, and I suspect many others don't either. All 3 are solutions for hosting something in the cloud, but what that is is many things, and writing code in a tiny portion of that.

The creation of these collectives, and the wealth of fundamentally off-topic tags that they contain, however, pushing unknowing users into a sense of confidence that a question about "How do I create a VM in Azure?" Is on topic; there's an azure-virtual-machine tag, and there's an entire portion of the site about Azure, it must be on-topic, right? Wrong.

In truth, when using the site for the tags I participate in, if I see a question that is part of one of those collectives it is, in fact, not on-topic but about Administration, and very likely should be asked on DBA. That the collective exists, and that such questions get answers, however, turns things grey. When the topic about administration falls inside the collective these questions are less likely to be closed and more likely to be answered, but when it doesn't fall in the collective, then it's likely to be closed (or hopefully migrated if it's a good question).

This is confusing; we shouldn't be treating on-topicness differently because, previously, a company paid for their tags to be promoted in a collective. The other collectives, in my opinion, don't suffer this problem as they weren't sponsored, so their topics are far more focused and community driven. The sponsors ones were initially driven my a company, not the community, who had no idea how Stack Overflow worked.

I don't have a problem with the Microsoft Azure existing, but I do think that they are flawed right now and detrimental to the focus of topics Stack Overflow('s community) want. That they exist should not change what is on-topic within them, but not outside of them; that is confusing for members new and old. Though it would likely be (a lot) of work (from the community), I feel that a recretion of them, from the ground up, would be beneficial for them. That can, from Stack Overflow's side, has a discussion for what should be included. If we have tags (and there are lot of them in those collectives), that could result in easily off-topic questions (or just fundamentally are), perhaps we could help direct users to an appropriate site in the community before they post something off-topic for Stack Overflow.

Most of my Stack Overflow focus has been in the container space, especially kubernetes. There's a range of questions here. These range from clearly on-topic questions:

I'm developing a Kubernetes operator in Go using the Kubernetes SDK, and ...

I'm building an application in Visual Studio Code, and want to use its devcontainers feature so my colleagues and I use consistent tools...

To clearly off-topic ones:

I'm deploying a Kubernetes cluster on bare-metal machines, and...

I'm trying to configure the Docker macvlan driver to give my containers individual network-accessible IP addresses, and...

In between these, there is still a big container-oriented gray area. Let's take the Kubernetes API: a Go program using the Kubernetes API is on-topic; questions about implementing configuration-driven logic in a Helm chart with loops and conditionals are on-topic; so is the core API itself on-topic, or does it become "not programming" because you're writing YAML? Or are Docker Compose questions on-topic if you're packaging your own application, but off-topic if you're running somebody else's?

I've felt like the community has gotten stricter here in maybe the past two years. Of the 5 highest-scored docker questions, two are closed as off-topic. Older questions about "how do I install Docker" tend to have gotten through; newer questions about "how do I install Docker Desktop" without a clear programming focus tend to be closed. Questions around basic Docker container setup could be programming-related or not (and I've definitely answered enough of them) but it seems like they garner close votes much sooner than they used to.

To pick a specific example, From inside of a Docker container, how do I connect to the localhost of the machine? is the canonical question for connecting from a container program to a non-container database on the same machine. It has a score over 3,500, and at the same time it was closed as "off topic" in April of 2022. I use it as a dupehammer target fairly regularly. Of the various personae who could use Docker, a developer seems fairly likely to have this specific setup (an ordinary user would run an all-in-one Compose file; a system administrator would frequently target an external database). But it's not "a programming tool" or "a programming problem" by the current standards.

One thing I think would be beneficial here would be clearer direction on what should happen to these container-oriented questions. "Installing Kubernetes" seems like it belongs on Server Fault, many CI-oriented questions make sense on DevOps.SE, maybe "running Jupyter Lab" is right on Super User. There's not a lot of guideposts in Stack Overflow suggesting that somewhere else might be better to ask. The volunteer-moderator tools have limited options for migrating questions, and text that discourages suggesting migration to an unfamiliar site.

There's occasional mentions of tag-tips. It'd also help some to warn that, for example, kubeadm questions (about installing Kubernetes clusters) are almost always off-topic.

It would help if we used less derogatory terms than "subjective" and "opinion-based". I've very often seen deep probing questions closed with these labels, which is a great shame as they can often yield important insights. A question, for example, about the merits of different ways of organizing Java unit tests, is likely to attract a range of answers based on different people's project experience, and dismissing these answers as subjective or opinion-based seems entirely wrong to me.

We actually encourage people to ask shallow questions rather than deep questions. We reward people for asking "why is my bubble-sort code running slowly" as opposed to "how should I go about choosing the best sorting algorithm for this particular problem". That can't be right. We should do more to encourage thoughtfulness and awareness of the full range of the engineering decision-making process.

The fact that a question doesn't have a simple objective tick-the-box right-or-wrong answer doesn't make it "a matter of opinion" as if all opinions are equally valid. There is good engineering and bad engineering, there are designs that are a good fit to the requirements and designs that are a bad fit. It's not all about debugging the code. Just because design and process questions aren't black-and-white doesn't mean we shouldn't be talking about them.

(And while I'm about it, reading other answers here and the reaction to them, questions like this deserve more subtle feedback than anonymous upvotes and downvotes. If people disagree, we need to know why.)

These might also be questions that seek to compare two or more items (e.g. looking for pros and cons of X and Y).

To me, these sound acceptable and on-topic. But I also know that not everybody believes that.

Comparing things is done extremely often by programmers. For example: "Do I use an array or a list?" there is no generic answer to this, as it depends on circumstances. But "What are the pros and cons of arrays and lists?" is as close as you can get to one. Describing two items in a comparable manner allows one to choose what they want based on their situation.

However, I know people close vote these kinds of questions. This may be a grey area.

Have perceptions or rules around closing questions for shifted over time? Looking back on your experience with Stack Overflow, do you feel the boundaries of what’s allowed have changed, perhaps gotten stricter or more lenient?

Yes, they've absolutely changed. I wouldn't say they've shifted, though. Rather, they've become harder to pin down.

Fifteen or sixteen years ago, I understood what was and wasn't on-topic. Over time, there were some adjustments (and even flip-flops on whether or not find-my-bug questions belong here). But the changes were usually communicated. Overall application of the rules seemed rather consistent. Today, however, I have almost no clue as to what somebody's going to consider on-topic or off.

As the volume grew, so did the pressure to close bad questions grew. In response, SO reduced the number of votes needed to close a question. I believe that contributed greatly to the inconsistency.

The list of reasons available when voting to close a question has been reduced and become less specific over the years. There are times I vote to close a question, knowing exactly which radio button I would have used in the past, and am presented with a non-orthogonal list of vague, generalities to choose from.

We also don't seem to care about communicating an accurate close reason. Suppose a question gets three close votes: one says it's off-topic, one says it's a duplicate, and a third says it lacks focus. It's possible all of those are true. (But if it's off-topic and lacks focus the arguably the duplicate question should have been closed as well.) Let's suppose only one of those close reasons is correct. If it's not actually a duplicate, but the box at the top of the question says it is, that doesn't help the person get their question answered, it doesn't help them understand how they could have written a better question, and it doesn't help anyone else who encounters it reinforce or correct their understanding of what's in scope and what isn't.

Back when I tried to do my part going through the review queues, I discovered that there was no incentive to correct the close reasons. If a bad question was closed, the reviewer was suppose to agree with the decision, even if the justification was factually incorrect. (This was one of two reasons why I gave up on the review queues).

It's immensely frustrating to find a great question, spend 20 minutes composing a solid answer, only to have the answer rejected because the question was closed as off-topic by an overworked moderator or hi-rep contributor during that 20 minutes.

On the other hand, you can commonly find a minutes-old question that's absolutely 100% unambiguously off-topic, only to find that it has already garnered three to five mediocre answers (including one that's been accepted) because rep-hunters would rather spend 30 seconds writing a half-assed answer than vote to close it. Fifteen years ago, that happened far less.

I haven't seen any attempts to help re-align the collective understanding of what's on- and off-topic. All of the changes I've seen over the years seem to blur the reasons and reduce the feedback.

One of this things that might need to be reviewed is the idea of having content and a model that fits all programmers needs, as the needs and the perception of what is useful is not the same for

• Full developers Vs Scripters / Low-Code users/developers
• Beginners Vs Advanced programmers
• Programmers with Computer Science background Vs People with programming only background
• Local Vs cloud based programmers

There are legitimate technical questions regarding things like program design and coding style. Contrary to popular belief, for established technologies this is not subjective at all. Good program design is pretty universal across all languages and there are many established best practices with a overwhelming consensus behind them.

For example - "code should be divided in small autonomous units that only concern themselves with their designated task and not with unrelated parts". This is true in any programming language and has been that way since the 1960s - it is not subjective.

And yet specific and narrow design questions are closed as "subjective". Frankly: this is a rotten site culture problem on SO. A lot of design or coding style questions are not subjective and thinking otherwise originates from the similarly rotten belief that "programming is art" - it is not, it is a craft or even an engineering discipline. In engineering, nothing can be allowed to stay subjective.

Furthermore, program design is an important topic, perhaps the most important topic of all of programming. Because with bad or non-existing design, all programs ultimately fail. Dismissing the most important questions the professional programmers might ever have, while keeping completely uninteresting 'missing semicolon' syntax questions, is rotten site culture.

Another example: the style question "What space width should I indent my code with" is perfectly answerable for any given programming language and it is not subjective. For the majority of languages, the answer is "You can do either 2 spaces or 4 spaces" and that is the only correct answer. ("But I like 8 spaces!" Tough luck, this is engineering, fall in line.) But if the question is "should I indent by 2 or 4 spaces" then yeah it is suddenly subjective because there are two industry de facto standards and no consensus over which is better.

So there was never a need to categorically close all questions like this as "primarily opinion-based". They could be evaluated as such on case-to-case basis, since there is all manner of different categorizes of supposedly "subjective/exploratory questions":

• "What space width should I indent my code with in language x" - perfectly answerable and on-topic on Q&A. Or if we for some reason insist on keeping our rotten site culture, migrate to softwareengineering.stackexchange.com (...another site with similar rotten culture...).
• "What space width should I indent my code with" (no language stated) - too broad, close.
• "Should I indent by 2 or 4 spaces?" - primarily opinion-based, close.
• "Should I indent code?" - too broad and open-ended, close. (Yes, there are stupid questions...)
• "How do you like to indent your code?" - too chatty/asking for opinions. Suitable for Discussions or chat but not Q&A.
• "What coding style is best?" - too broad and open-ended, close. Anything about "best" without defining best should be closed, not suitable anywhere.
• "What coding styles exist and are commonly used in language x?" - too broad for Q&A ("list question") but perhaps suitable for Discussions.
• "What coding style was used in project x?" If the code was never public - close since it isn't answerable. Or otherwise it might be a better fit for retrocomputing.stackexchange.com or Discussions/chat.

And so on, I can keep making examples and there's no universal truth for how to deal with these. Treat them individually.