I am a developer of PETSc, an open source library of parallel solvers and related tools. We currently have a great deal of redundancy on our mailing lists and are looking for ways to provide better answers with less effort. Our FAQ answers a lot of the common questions, but users rarely search this or the mailing lists before asking questions by email. Note that many of our FAQ questions are clearly relevant to a larger scientific computing audience, for example

What about more package-specific questions?

Here is an example of a reasonably non-duplicative petsc-users thread that I think could become a good question:

To what extent would package-specific user support questions be allowed here? Assume that they are tagged [petsc] and that duplicates are managed by us. (We do the work anyway, but I think it would be more valuable in this setting than restricted to mailing lists.)

This is a similar situation to e.g. Trello on Webapps (see meta discussions here and here). What emerged out of that case is the idea that bug reports and feature requests are definitely off topic, but usage questions are generally on topic, as long as you allow them to arise naturally. So I suppose the same could apply for PETSc. I certainly don't see anything wrong with having a tag for a reasonably popular software package. In fact, I think it limits the usefulness of the site if people are forbidden from asking questions about usage of specific programs.

There is also a distinction to draw between (1) questions about the actual application of PETSc (or any other package) to scientific computing tasks and (2) technical questions about the software, such as the one about the configure option. The latter category is not really about scientific computing, except for the fact that it happens to be specific to a software package that is used for scientific computing, so one could make the case that it doesn't really belong on this site. But that will be decided elsewhere on meta.

  • 2
    Also, can't people with high rep edit questions to make them more general but equally useful to the person who asked them? – Aesin Nov 30 '11 at 12:50
  • 1
    I agree with this answer very much. I, for example, work extensively in Matlab, and while I believe that programming questions should go to SO, it would be arbitrarily limiting if I couldn't ask a usage question on here. I'm imagining some question like "Is <matlab function> numerically {stable/efficient} when used in my implementation of <algorithm>?" – Colin K Nov 30 '11 at 17:16

I am definitely for the first category. I think for a package like PETSc with thousands of users, the latter questions are also probably appropriate, but it would be nice to have a flag or section for them. I am not sure how to handle the more open ended questions like the last one, although at least one of these has already been asked so maybe its alright.

I think the rule for specific packages should be that of Supply and Demand - if there's a tag for it, and questions are being answered, why not?

  • 3
    Yes, but the question is: should we answer such questions here. So this answer is a bit circular. – David Ketcheson Nov 30 '11 at 14:54
  1. Questions about software or libraries should be appropriately tagged.

  2. Questions about the math/physics involved in a calculation should be answered as points of clarification.

    For example: "What approximations are being made when I do a CASSCF calculation in Gaussian 03?", is a question that clarifies the technique but does not address the approximation(s).

  3. Questions about the validity of an approximation should be a community based judgment call.

    3a. They should be voted on by the community through comments and comment up-votes:

    Is it essential that these questions are answered here?

    Are there two questions here, one about Computational Science and one about math/physics, and should they be split?

    3b. If it is voted that the question is more math/physics based, then the question should be migrated to the corresponding SE site by a moderator.

    3b. If it is voted that the question is more math/physics based, the community should suggest, through comments, how the questions might be improved to fit with in the purview of this site.

  4. Questions about compilation errors, runtime errors, compilation flags... should be migrated to a mailing list or reworded and migrated to stackoverflow.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .