Computer-algebra tag?

Question

Are questions about implementation of specific tasks in specific computer algebra systems on-topic here?

If so, can we create a computer-algebra tag?

Context: I have been feeling around on the StackExchange network for an appropriate place to ask nuts-and-bolts questions about how to implement specific mathematical tasks in the computer algebra system Magma. It was suggested to me at meta.math.SE by user quid to try this site, and just on the basis of the site description this seemed like the right place, as these questions are certainly about "computational methods used in technical disciplines". At the time that I asked my first question, there did not exist a computer-algebra tag or a magma-cas tag. A magma-cas tag has subsequently been created by user hardmath.

Answer 1

I'm tempted to say it depends. If the question is "What software or library can I use to implement task X?", I'd say yes, it's on-topic. If the question is, "How do I use software package/library Y to implement task X?", it's probably off-topic. Here's why:

For a while, we've had questions about software libraries on the site. For instance, in the early going, we had a fair number of core PETSc developers on the site actively contributing, so there were several PETSc questions that would get asked, and they would be answered by very knowledgeable people. That's sort of died out.

Similarly, we tried an experiment with FEniCS where they would direct users to ask questions on SciComp, and FEniCS devs would answer them. That also didn't go well, because we seemed to experience extremes: the FEniCS questions would swamp out other questions on the site during FEniCS-heavy days, or the FEniCS questions would get buried in a sea of general computational science questions when fewer FEniCS questions were asked. I think the feeling from that experiment on both sides was that both communities were poorly served.

So, the tendency is to close these "how do I use this library" of questions and refer them to software mailing lists, unless the software in question is so widely used (e.g., LAPACK, BLAS) that someone in the general computational science community is likely to give a sufficient answer. I think long-time users in the community have found that it dilutes the focus on software developers, because now they have to look at their mailing list and this site, and if they're not interested in answering general computational science questions, it's less time they spend doing their core work maintaining their software. I think I initially felt that it would help boost traffic to the list to get discussion on many different software packages, but now I've come to feel that the mailing list model is probably better for this sort of discussion.