What do we do about software recommendation questions?

Question

In my short time at AskUbuntu, I have seen software recommendation questions ("What IDEs are available?", "What rich text editors are available?" [that one was actually asked, answered and upvoted quite recently], etc.) pop up fairly regularly. AU already has quite a few of such questions with extensive answers. The classic example would be What is the difference between the vim plugin managers? (perhaps rephrased as: What plugin managers are available?).

Based on that very limited experience, I predict we will see a few of those here as well once the site goes public.

Instead of closing those as OT/Too Broad, can we:

• allow/make a few of these now, curate the answers
• close/lock them when we go public, agree to not delete them, and
• close new questions as duplicates when the public starts posting them.

A few ideas:

• resources for learning
• handling Tex/Latex
• turning vim in to IDE

Answer 1

I'll give my perspective as a long-time Stack Exchange user over diverse topics, and moderator on Software Recommendations.

Recommendation questions are problematic because they tend to degenerate into popularity contests. Popularity contests do not produce useful answers: at best, the top answer ends up being one that pushes the most popular tool, regardless of whether it is suitable for the purpose set out in the question.

Nonetheless, recommendation questions can work well on Stack Exchange. The key is that a good recommendation question is not “what's the best …?” but “what's the best … to do X?”. If an answer recommends an otherwise excellent tool that does not do X, downvote it.

Software Recommendations applies stringent moderation guidelines to both questions and answers to keep out questions that are missing the “to do X” bit and answers that don't explain how this oh so marvelous tool can do X. This works pretty well. What works less well on SR is that there are few experts in any given subject. The best place to ask about a Vim plugin for a particular task is Vim SE, not Software Recommendations SE.

So, don't get hung up on the fact that a question asks what tool to use for a job. It's still a question about how to do that job. The fact that the reasonable way to do the job is to use an existing tool does not make it a bad question. Some Stack Exchange sites throw the baby out with the bathwater and shoot at any sight of a question that might be asking about existing tools. Don't imitate them, instead imitate the other sites that understand that the answer to “how do I do X?” might well be “use this tool”. Do, however, apply normal quality control. A question that asks “what's the best X?” may well need to be closed as unclear, with a comment asking “best X to do what?”.

---

What is the difference between the vim plugin managers? in its original form was problematic because it lacked the “to do what” part. It's been rephrased as asking for a comparison rather than asking for a recommendation, which is fine: it asks to be taught how to fish, when the original was asking for a fish.

---

A few comments on your proposals:

• allow/make a few of these now, curate the answers
• close/lock them when we go public, agree to not delete them, and
• close new questions as duplicates when the public starts posting them.

No. Locked questions are absolutely useless. What do you do when a newcomer appears on the scene? What's so special about the private beta that some questions can be asked now but not later? If a question is worth asking now, it's worth asking later.

“Canonical questions” — questions that are relatively generic, and have answers which the community has spent a lot of effort getting just right — are definitely useful. They don't tend to be asked during the private beta though: they are frequently asked questions. Usually, a canonical question is posted after dozens of minor variants of the same subject have been posted. Once the canonical question exists, all these minor variants can be closed as a duplicate of it. (Take care not to close “exotic” variants not covered by generic answers though.) Canonical questions are never locked — they need to have top-notch answers, so being able to edit them is critical.

A few ideas:

• resources for learning
• handling Tex/Latex
• turning vim in to IDE

All of these are far too broad topics for a Stack Exchange question. You could write at least whole book chapters on each of these. It's telling that all of these lack a “… to do X” part.

Answer 2

First is the underlying problem. The root of any "recommendation question" in general, if we pretend the construct exists, is some problem that needs to be solved. A software is sought out because there's some task that they need software to accomplish. If the underlying problem itself is something that is interesting or useful, especially something that the community itself thinks is useful, that's important to keep.

This is usually the point where I'd tell people to just "fix the verbiage". Change things from "Is there a plugin that does X" to "How do I do X?" That way it keeps the question open to both direct solutions as well as software/plugin/etc. recommendations. It is, in many ways, a humongous formality to do as such though. If the community as a whole can learn to recognize when there's a useful problem behind the question as asked, then less attention can be directed at the raw verbiage used to convey that problem. All the same, it still does make things easier for everyone in the long run to fix up the way a question is written.

There isn't always a strong underlying problem. And that's basically where questions break down and become off-topic or otherwise unsuitable for the site. For example, asking how to do something particularly basic, instead of asking for software that accomplishes that basic task, is probably overly broad. But that actually wouldn't change the broadness either way it is phrased, and that's key. Most any poor-form "recommendation question" can probably be identified as not belonging on the site for reasons unrelated to it being labelled as a "recommendation question".

---

One final consideration though, is that value to the community is important. Sometimes, there actually is like one or two kinds of questions that are actually really useful to people in that field. I don't know much myself on this particular site - I use Cream, a "modernized GUI of Vim", and don't actually delve into the deeper functionality of the system. So I can't say whether or not there does exist some exceptionally useful canonical question. But, if there is one such question, not creating a question class that would ban it would be pretty helpful to avoid weird conflicts of keeping such a question while closing other similar ones by policy. As mentioned earlier, most any of these can be handled without needing to make an extra rule about this system.

Answer 3

A possible approach would involve restructuring the question to improve the nature of its answer.

For example, if someone were to ask "What plugin managers are available?", a strong restructuring would be to the question "How do I install plugins in Vim?", for which answers would include information about manually installing packages, as well as answers for how to use a plugin manager.

This could apply to many, if not all, "What are my choices for [topic]?" or "What's the best way to [task]?" questions.