Is 'don't do this' a valid answer to a question looking for input on how to do something? [duplicate]

Question

Submit and Reset button, what's the best order? asked on UX has as the highest-upvoted answer, something that does not address the question at all: instead it just points out that this is generally a bad practice.

The asker already commented that its not applicable in his case as he can't easily dispense of a Reset action, since then I've seen the upvotes on this answer go from ~10 to 30+.

If the intent is to point out bad practices, that can also be done by actually answering the question and mentioning the drawbacks of its premise (instead of ignoring the actual question asked). Should such replies be encouraged as actually answering the question? Personally I feel this would be more suitable as a comment on the question instead.

Another issue is that the later upvoters probably ignored or never read the asker's comment, but that's another discussion on whether useful info can be left in comments or not :)

---

2 mins after posting, I saw Is "Don't do it" a valid answer? in the sidebar. This is almost a duplicate, but has the added proviso that the 'don't do it' answer did get replied to by the asker promptly that yes, he had considered this aspect but was going to do so anyway.

Answer 1

Yes, if it's a good answer in its own right. It needs to include actual expertise or research or facts saying why not to do something, and ideally suggest an alternative.

Sometimes, askers say what they're trying to do and what they're stuck on, then an answerer tells them to try an entirely different approach. I would view a good "don't do this" answer as in this class of answers.

I'll also note that sometimes these answers can be blunt and snarky. I don't think that's constructive - the OP didn't really learn anything besides the fact that at least one person on the internet thinks she didn't do that, which she probably knew before posting a question. These should be constructively downvoted, with a comment explaining that this could be a good answer if it were educational instead of curt.

Answer 2

If the question is:

Should I post my question without first looking at the recommendations of questions that may already have my answer, which are populated immediately after focusing away from the title?

Then the answer is don't do it! Because #2 in that list should have been this duplicate with a very similar title and plenty of discussion:

In all seriousness, just because a user asks for the best route from San Francisco to Los Angeles - by tricycle - does not mean it's a well-founded question or that anyone is well served by strictly conforming to some rule that we must answer the question asked and only the question asked.

Now, I don't think there's any place for an answer that just says:

Don't do that. It's stupid.

However, an answer that doesn't answer the question, but offers an alternative is probably useful, since most reasonable people would not every try or even condone riding a tricycle between these cities:

I wouldn't advise riding your tricycle on any main roads anywhere, never mind for that distance. I would suggest taking the bus and leaving your tricycle behind (or shipping it).

For a more relevant example, I've answered questions like, "Can can I optimize this MERGE command?" with an answer that basically said, "Just stop using MERGE." The key is that I gave a pretty lengthy explanation about why I was recommending against it. And they didn't accept my answer, but if you read through the self-accept you'll see that my advice did help steer them away from MERGE and to a better, more efficient solution. If I follow some rule that says I can't tell someone not to do what they're trying to do, my only viable answer would have been to work with their MERGE command and try to optimize it.

Answer 3

This is my own opinion on my question, as this is a discussion - of course I do want feedback from the community as well.

One of my objections to doing this is mentioned in an answer to the linked question, and this other answer covers why I think questions do need some kind of actual answer.