When is a question/answer too program-specific, and what is the policy on redirecting to other forums?

Question

The question came up in this thread about an OpenFOAM-problem: user asked a question on the appropriate forum where he didn't get an answer too (or didn't find it in the first place) because he didn't know that the thing he was looking for is called "baffles" in OpenFOAM (so it is a pure nomenclature question).

My answer was the nomenclature and then referred him back to the OpenFOAM message board which sparked a little discussion.

This question approaches the topic from "the other side": how should Scicomp act for software that has a "living" community? I would suggest that it doesn't try to replace them but give overview answers and then refer to that community.

I love the StackExchange-sites especially for stuff where

• the answer is unlikely to change (mathematics and general programming)
• the community (and the available resources) are too large and it is not clear where to go for information ("How do I do .... in Python" when one doesn't want to subscribe to mailing-list for that answer and looking through archives yields no "converged" answer)

In my opinion for the the concrete example of OpenFOAM I'd prefer if only questions like

• "Is it possible in OF to ..." (capabilities)
• "How do I ..." (general directions)

where answered here in a general way but for the more concrete advice (calling utilities, program examples) people where directed to the MessageBoard. Especially as concrete usage is likely to change in subsequent versions and as I understand it the idea behind the StackExchange-stuff is that the answers to a question should converge to a form that doesn't have to be maintained later. The MessageBoard is "living", quality there is quite good and I'd prefer to have all relevant answers in one place.

What is the general consensus on this?

Does SciComp even strive to replace existing forums?

Answer 1

SciComp does not intend to replace existing forum. I see it like StackOverflow: While at first you might wonder why any Java developer should care about Python questions (and vice versa), there is so much common ground that they can still provide valuable hints on a problem even without being able of expressing that in the specific programming language. And just the same way in that specific question of mine it could have happened that someone who does not use OpenFOAM might still have known "the things you're looking for might be called baffles". And I hope that nomenclature itself is not subject to change too often that an annual edit can't fix it... Then again, a tag can indicate the major version, e.g. python2 vs python3.

My personal point of view (and therefore not necessarily consensus) is that a forum is more adequate for very individual problems ("Why does my simulation segfault?"), where one e.g. posts an entire case and with the help of other users iterates on finding the cause of trouble (and maybe someone else having a similar problem picks up the thread, although this unfortunately sometimes leads to quite a change of topic...), while at SE the problems are of a more generic nature ("Which boundary conditions should one use for....?") such that others with similar questions do not to interact any further but reading the question and up/downvoting a bit.

Answer 2

When I get back to my computer, I'll edit this post to include links.

I've been thinking about this question for a while. The way I see it, SciComp is intended to be a cross-disciplinary forum. While it might be detrimental to the site to have it be swamped by package-specific questions, we're nowhere near that point yet. There have been a number of PETSc, BLAS, and LAPACK questions that have been useful to people and certainly on-topic.

At the same time, SciComp isn't intended to replace software-package-specific forums, where package-specific questions are likely to be answered by people at least as knowledgeable about the software, if not more so. The audiences for each forum are almost certainly different. If you ask about how to do something in a software package that lacks the capability you're looking for, you're probably more likely to find someone on SciComp that can point you to software packages that do have the capability you're looking for, whereas on the package-specific board, you're probably more likely to get a response telling you that what you want to do in that software is impossible. The two fora have different strengths.