There has been several discussions in several meta-posts on this. I think we should settle the issue once and for all.
Do you think programming questions should be on-topic or not? What kind of programming questions can be on-topic?
Please be specific about what you mean by "programming questions", and give examples of the type of questions you are discussing.
I think the programming questions should be off-topic for several reasons:
By a programming question I mean a question about code in a particular programming language and the point of asking for the code is not the programming language itself. A question about a feature of a programming language can be on-topic, e.g.
Questions which are asking for algorithms should not require a prespecified language for the answer and an answer written in pseudo-code should be acceptable.
Questions asking about code in a particular programming language that are not about the programming language itself are off-topic. Typical examples are:
You can replace "C" with another programming language, and replace "multiply to numbers" by another computational problem.
Above I have argued that there is not a need to have programming questions here, and based on other factors like the intended scope of the site, it is better to leave them out. But can having programming questions lead to bad things for this site?
I think so: as demonstrated by half a million users of SO, there is a HUGE community of programmers out there with HUGE number of questions, much larger than the computer science questions we will get. By allowing programming questions we face the risk of computer science questions being a negligible portion of the site and turn the site into another version of SO, which would defeat the purpose of the site as a place for computer science questions (and computer scientists would leave also for the same reason that they are not active as computer scientists on SO). Moreover, since SO works very well for programming questions, we are likely to get question with lower quality than SO.
Define programming as the act of writing a program. A programming issue is some issue that prevents you from writing a particular program. An algorithmic issue is a programming issue whose cause lies in the programmer's understanding or knowledge of some algorithm (to the point of not knowing the algorithm at all). I won't attempt to define what is an algorithm and what isn't.
An example of programming is me, trying to get my LR(1) parser generator working before Thursday. An example of a programming issue is the C# compiler crashing when I tried to compile some of my generated C# code that initialized my tables (the tables and therefore the code file was too big). An example of an algorithmic issue is me not knowing how a shift-reduce conflict should be solved if precedences are given and I have multiple conflicts in the same state on the same lookahead.
All programming issues that are not algorithmic issues should be off-topic.
There's really only one good reason for this: a huge part of SO would otherwise become on-topic for this site. Overlap is a good reason to declare a topic off-topic: the reason this site has only come into being a month ago and not years ago is that there was perceived potential overlap with other sites. Feeding that perception by declaring half of SO on-topic is a very bad idea.
The second reason is that programming issues that are not algorithmic issues don't really belong to computer science: a computer scientist shouldn't be interested in all the nitty gritty details of the Real World, as signified by big-oh notation and abstract models of computation. These issues are caused by irrelevant details that can be overcome, and we should therefore not be interested in them from a true Computer Science standpoint.
Note that these 'details' I'm talking about may be very complicated and hard to overcome, but there's no interesting theoretical background to them: details that do have such a background, like IO-efficiency, numerical analysis (finite precision), human-machine interaction, have been formalized and studied as Computer Science, and should therefore be on-topic.
Now, assuming that we indeed make this off-topic, there are two further points to this decision. We need to decide where the line lies, and how to communicate this line.
My proposed criterion is what I already said above: any issue that is algorithmic in nature should be on-topic, everything else is off-topic. It shouldn't matter what programming language is used to communicate the algorithm (issue) in question.
A clearly off-topic question would then be "What does this piece of Perl code look like in Python?". An on-topic question would be "I'm trying to implement an algorithm that uses hashtables in functional programming language X, but it doesn't have hashtables, what should I do?". An off-topic question would be "What's wrong with this mergesort implementation" (assuming the problem is, say, an off-by-one error somewhere).
On how to communicate this line, I'd suggest doing two things: have a vague, understandable rule-of-thumb that errs on the side of on-topicness, along with the exact criterion above. With any luck, the people that aren't sure what their problem is decide to try their luck based on the vague rule-of-thumb, while we still have the exact criterion that empowers us to close questions that are off-topic, without the possibility of a discussion whether that closure was warranted based on the rules.
Well, I seem to be the one instigating all of this, so the least I can do is to provide an answer and let the community treat it as it well.
Short answer: It depends on the programming question.
Long answer: I believe that the best solution is to:
There are several reasons for which I hold this belief:
Declaring "computer programming" as off-topic unfairly and unproductively excludes a large number of potentially interesting questions and the users who might ask them, receive answers from a computer-science perspective, benefitting both themselves and the community.
Declaring "computer programming" as on-topic provides the community with no charter, or mandate, or moral authority to exclude a question only on the basis of its being a programming question, which could make it impossible to meaningfully distinguish this site's scope from a variety of other sites' (Stack Overflow, Programmers, etc.)
The community's reserving the right, without providing a guarantee, to close programming questions as off-topic is both more inclusive than the former solution, and does more to appropriately limit the scope than the latter solution. Where it is weaker than the others is that it does not provide a clear answer to the question of whether any given question will be on-topic.
I believe it is better to be inconclusive all the time, and allow specific circumstances inform the community's decision, than to uniformly (and, in some cases, possibly incorrectly) influence the community's behavior, when it comes to passing judgment on questionable content. If the community doesn't agree that the relationship between programming and computer science is questionable, or debatable, or controversial, that is a separate discussion... and one which could be impossible to satisfactorily resolve.
Alternative
Alternatively, or in addition to the above, the FAQ can attempt to provide a working, or practical, definition of the site's scope, so long as the motivations are made clear in the FAQ (e.g., too much overlap with Stack Overflow, or too much potential for poorly-placed questions, etc.)... so long as the policy which informs and guides the community is accurate (in terms of representing true motivations).
Examples of programming questions which could be considered on-topic
Given this language or formalism implementation of algorithm or process, what is the meaning of part of implementation / what does part of implementation do / why is part of implementation needed / can part of implementation be replaced with part of implementation?
Reason: Such questions require (a) the ability to read and understand the syntax and semantics of the language or formalism being used, which might come to bear on any correct answer to the question (of course, it might not!); (b) the ability to read and understand the a*algorithm or process* which the implementation encodes, and what effects on correctness/efficiency/etc. changes to the algorithm or process would have; (c) the ability to relate the ideas of algorithm or process to language or formalism, which is of fundamental importance in computability and complexity (i.e., can you do X in Y, and if so, can you do X in Y with complexity Z?)
More generally,
Does definition of computational process or structure, expressed in language or formalism, satisfy the property well-defined property for the indicated language or formalism?
Reason: Computer science provides unique tools for analysis and understanding of computational artifacts, allowing (in many, but not necessarily all, cases) for clearly reasoning about characteristics of those artifacts. The language or formalism employed to express those artifacts should not be a primary concern, so long as the the communication to the community is successful, and the artifact/language (or appropriate models thereof) are subject to the tools of computer science.
Note that nobody, including me, is advocating allowing questions that ask for an implementation in a programming language. Those questions should be closed, but for other reasons. Questions which do not show any effort to understand the problem can be closed, for other reasons. Questions which are argumentative or which use a significantly non-standard or incomprehensible notation can be closed, but for other reasons. Otherwise perfectly acceptable questions, for which this community fails to provide a reasonable answer, can be migrated to appropriate sites. It's this last point that presents the only difficulty with my solution: unless SE provides automated means of identifying such questions, it could mean a lot of work for moderators, or questions languishing for a while. Whether this problem would be serious or not is unknown; and in any event, I believe the community could find a way to fix it.
