"No" is a valid answer to "is this possible?". Sometimes it can be quite useful. A recent example: How to ban all connections to .se and .ru in the hosts.deny file
- Change format of syscall event trace output to ftrace
EDIT: In the two links above, you can currently see a "no" answer which was deleted, and converted into a comment. But looking at When to "Comment", and when to "Answer"? , I would think it fits better as an answer, and does not fit well as a comment.
This type of answer might be most useful when worded "It is not possible in version 42. The current features (or architecture) are exhaustively documented <here>, and none of them will let you do this."
I think there is no written rule that requires that wording. I can imagine a moderator wanting to avoid the "this answer is not useful" scoring process etc, for an inexperienced StackExchange user. But I am not convinced this is the best approach. As an example of an alternative, I saw another short, new answer today, which received a comment response "Do you have any references for this?"
In this particular case, an edit like "citation: I am the first name listed in the
MAINTAINERS document for this code" could make the answer more useful. The author has since attempted to provide this information.
This answer is not simply a pedantic mis-understanding of a common idiom. It would still apply when question posts are worded "I want to do X, and I need to do it using tool Y because reason Z". The correct answer may be "tool Y cannot do X".
The caveat here is that when you cannot tell what reason Z is, it can be unclear what the intended question is. Does the asker require answers to use tool Y, or did they just assume that tool Y would have a good way to do X? That is what we have question comments for. And ultimately, the ability to put a question on hold. I.e. if a post is not a good fit for this site's definition of a question: a post that can have a definitive answer.