What's potentially wrong with questions about alpha software?
If they turn out to be a bug or based on an unstable API, they are too localized, as they will be irrelevant soon. However, just as easily, the user could be using the API incorrectly, so it's not really clear which are "too localized" problems and which are legitimate SO questions. Unless as a community we take the stance that it is not okay to ask questions about immature software, we must find a way for at least some of these questions to be permitted.
Furthermore, note in the case of something like Android, it has a mature release cycle, and its owners endow it with a well-defined alpha stage. But SO supports many smaller programming technologies and communities, some of which may have more bugs in mature products, some of which may have perpetually unstable APIs for better or for worse, so it's arbitrary to declare some software too young without any subtler criteria.
How do we handle questions when the OP doesn't have the expertise to know whether it's a good question?
This happens with the category "not constructive" often. Sometimes questions like "how do I do this simple thing?" happen that look constructive to the OP, but someone posts, "Wow there are a zillion ways to do this" and closes as not constructive. But these questions almost always indicate poor research effort, because the user could have done a Google search and discovered them (if there are a zillion, they're probably easy to google).
But there is a category of questions that are good when posted but closed shortly thereafter: exact duplicates.
- Some duplicates are bad questions. The user didn't research that much, finding the duplicate was trivial - it may even have been suggested to the user as a duplicate when posting. These usually receive downvotes and are quickly closed.
- Some duplicates are good questions. It may require real expertise to find the duplicate. For instance, at the "novice" end of programming questions, the OP may ask about "a function with changing number of inputs" and had no idea to search for "variable arguments to function," but someone with more expertise may know the magic words to use, and is able to use this expertise to find the duplicate. These questions may receive upvotes because they look like good, original questions until the expert steps in, then it is closed as a duplicate, and usually still has upvotes.
When should alpha software questions be accepted?
- In cases where it's unclear whether something is an alpha bug or legitimate user confusion, and is a high quality, well-researched question, they should be permitted.
- If you notice such a question and know it is a bug, or are capable of researching and determining this, please post this as an answer or as a comment with facts, reference, expertise etc. to back you up, and vote to close as Too Localized.
- Do not delete these questions. They serve as signposts to future visitors, much like duplicates do. They need to be alive for at least as long as the alpha period is open, so similarly confused users get to a "too localized" question instead of re-posting the question. Please tag them well so they can be cleaned up after alpha phase.
What about comments discouraging posting?
"It's alpha software, it's not supposed to be perfect!"
That's not an answer. Or helpful. Or constructive. Or really worth saying at all.
"Report the bug to Google, not here."
Yes! Do exactly that after we know it's a bug, then close this question as Too Localized. I'm going to take a wild guess and say most people posting that as a comment are just being unhelpful. Otherwise they would, you know, be helpful, by:
- linking you to a bug report
- explaining why they know it's a bug and not ignorance for how to use the API
- vote to close as too localized
So comments saying "report to Google, not here" with no other information are best taken as a suggestion that it may be a bug and not heeded further.