7
$\begingroup$

Recently we've been discussing here whether questions on numerical algorithms should be allowed when the context is a physics problem. The last time this was discussed on Meta seems to be here (a follow-up to this), which implies that such questions should be allowed. However, computational questions have historically beenoff topic according to the definition of our scope in the help center, and the last time I proposed changing that, the only response was to wait until the Computational Science site had taken off and reevaluate the decision then.

Well, now Computational Science is well established, and nevertheless it seems we are of the opinion that numerical algorithms in a physical context should be on topic here.

  • Is this assessment accurate? Does the community want questions on numerical algorithms in a physical context to be on topic for us? (this is just a last opportunity for anyone who disagrees to speak out)

  • How should we word the item that would need to be added to the Help Center's on-topic page under the "What topics can I ask about here?" heading to reflect this?

$\endgroup$
  • 5
    $\begingroup$ My opinion is that quality questions (that is, NOT homework-like) about simulation and numerical computation of physics don't hurt the site at all but could make the site more interesting / useful to some folks. $\endgroup$ – Brandon Enright Mar 7 '14 at 22:24
7
$\begingroup$

I would word the help-centre item like this.

In the list of "on topic" things:

  • Computational physics [[followed by a link to an exemplary question]]

In the list of "off topic" things:

  • Implementation details

    While computational physics is on topic, we are not a programming site. If your question is about implementing computational code - in particular, if it's about writing, compiling, debugging or optimising code, or about a specific language or library - then it is off topic. It may be suitable for Computational Science or Stack Overflow, however.

Finding a good exemplary question for computational physics is tricky due to them being off topic up to now. Perhaps we'd have to wait until a great one came along.

I've left out the issue of software questions (e.g. how to I do X in package Y), because it seems like a somewhat separate issue to me. We could always put an additional "software issues" item in the "off topic" list if we feel it's important enough to mention explicitly.

$\endgroup$
  • $\begingroup$ oh... now I feel silly. I completely missed that we don't have descriptions of the on-topic items. Still, it will be useful to have a description of what sorts of computational topics are not on topic here. So I like your general idea, but I don't think what you have here is specific enough. Perhaps we can come up with something that incorporates more of the specifics from tpg2114's answer? (for example) $\endgroup$ – David Z Mar 16 '14 at 3:22
  • 1
    $\begingroup$ @DavidZ I tried to cover everything tpg said, but in a similar length to the other items in that list. His "can't do" points were (1) implementation, (2) debugging, (3) optimisation, and (4) libraries/languages/compilers, all of which I managed to squeeze in there. We could be more specific, but then it would get longer. $\endgroup$ – Nathaniel Mar 16 '14 at 3:25
  • $\begingroup$ Hm... maybe it's just that your phrasing only seems to address questions about specific languages and libraries, thus implying that anything that's not about a specific language or library is on topic. In any case let's collect some more input and see what we can come up with. It's not bad, for sure. $\endgroup$ – David Z Mar 16 '14 at 3:33
  • $\begingroup$ @DavidZ that's a fair point. I've changed the wording to try and avoid that implication. $\endgroup$ – Nathaniel Mar 16 '14 at 3:42
  • 2
    $\begingroup$ @DavidZ I think the list I gave should probably be more of the basis of a FAQ to give examples and counter-examples in a bit more detail. In fact, we might want to be less specific in the help center and just say "While computational physics is on topic, we are not a programming site. If your question is about implementing computational code then it is off topic. It may be suitable for Computational Science or Stack Overflow, however." $\endgroup$ – tpg2114 Mar 16 '14 at 3:58
  • 1
    $\begingroup$ Mostly because we need to be careful about giving a list of things. Is it all-inclusive or just a subset of examples? People like to argue about silly things so it's better to not give an explicit list in the help-center itself. Plus that allows for more fluidity if desired in the FAQ -- easier to edit a meta post to grow/shrink the examples of what is good or bad than to reword the Help page. $\endgroup$ – tpg2114 Mar 16 '14 at 4:00
  • $\begingroup$ I totally agree with tpg's comments here. $\endgroup$ – Nathaniel Mar 16 '14 at 4:23
  • $\begingroup$ I've added this to the help center, since there don't seem to be any additional comments for a while. $\endgroup$ – David Z Mar 29 '14 at 19:12
11
$\begingroup$

I wholeheartedly support adding computational topics to our site, but where we draw the line needs to be very clear. Here is where I would draw it:

What we can't do

We cannot answer programming questions (How does one implement X method for Y equation?)

We cannot debug code (I programmed Crank-Nicholson but it's not working, what did I do wrong?)

We cannot optimize code (My matrix inversion routine used for Crank-Nicholson is running slow, how can I speed it up?)

We cannot advise on libraries/languages/compilers/etc (Should I use PETSc or Trillinos for my code? or Is Fortran better for scientific programming? or Should I use OpenMP or MPI to make my code parallel?)

What we can do

We can advise on numerical methods when tied to a physical problem and not a generic P/ODE (I am studying shock-turbulence interaction, is WENO or AUSM better? or What methods would one use to solve diffusion problems?)

We can discuss scheme quality/limitations (What makes the original Murman-Cole scheme not conservative in transonic potential flows?)

We can discuss scheme design (How does one determine if a scheme is conservative?) Note: This one may be controversial here as that might be better in SciComp.

The punchline

I say we include it but with the caveat that it needs to be tied to physical problems and not to programming. No implementation issues, no optimization, no CODE. Just concepts.

$\endgroup$
  • 1
    $\begingroup$ Given the remits of other currently active SE sites, I think this is a good place to draw the line. $\endgroup$ – Nathaniel Mar 8 '14 at 6:45
  • 2
    $\begingroup$ I agree. Further, I think we should work on any related tags like computational-physics so they list some of your caveats and limitations. We probably should also edit the simulation tag to make it clear where the difference is. With enough details in the tag wiki and tag synonyms we could probably keep things on-topic. $\endgroup$ – Brandon Enright Mar 8 '14 at 7:47
  • $\begingroup$ Perhaps a separate faq meta question is in order to describe what sorts of computational questions are allowed. But first I think we should come up with a quick blurb to put in the help center. $\endgroup$ – David Z Mar 8 '14 at 20:15
  • 1
    $\begingroup$ I completely agree with this answer. It was in this spirit that I posted this question, which I think was a great resource and a good fit for the site. I am disappointed that it was migrated, and I also note that no new answers have been added in Comp. Sci. $\endgroup$ – Emilio Pisanty Mar 8 '14 at 21:37
  • $\begingroup$ @EmilioPisanty I don't know how well this answers your question (in particular simplicity) but a new paper on your question just came out: sciencedirect.com/science/article/pii/S0010465514000678 $\endgroup$ – tpg2114 Mar 9 '14 at 23:55
  • $\begingroup$ @EmilioPisanty: I was disappointed that it was moved (along with the 140 rep). Even more so when I saw that it was downvoted 3 times within an hour by a bunch of idiots who can't understand the word "effectively." $\endgroup$ – Kyle Kanos Mar 11 '14 at 16:12
  • $\begingroup$ @Kyle Wow, I hadn't seen that. Sheesh! A whole lot of good done to that thread by that migration. $\endgroup$ – Emilio Pisanty Mar 11 '14 at 17:14
  • 1
    $\begingroup$ @DavidZ haven't we twice before had the discussion about which sorts of computational questions are allowed? The consensus was the same each time: essentially all questions about computational physics are allowed, but implementation details are not. This is also what tpg said here. I don't see a great need to have that discussion a third time. $\endgroup$ – Nathaniel Mar 15 '14 at 16:37
  • 1
    $\begingroup$ (the past meta discussions on the subject are meta.physics.stackexchange.com/questions/43/… and meta.physics.stackexchange.com/questions/2898/…) $\endgroup$ – Nathaniel Mar 15 '14 at 16:40
  • $\begingroup$ (also note that unless I'm missing something, questions along the lines of @EmilioPisanty's one discussed above have never been off topic.) $\endgroup$ – Nathaniel Mar 15 '14 at 16:41
  • 1
    $\begingroup$ @Nathaniel I think the issue on the previous discussions is they are old -- 2010 and 2012 -- and people were flagging Emilio's question for migration which might indicate a change in attitude toward the on-topicness. But based on the discussion (or lack thereof) here, the attitude hasn't changed and those questions should remain. So it's okay to revisit in a "Do people still feel this way" conversation since it's been awhile and there was disagreement about a question. $\endgroup$ – tpg2114 Mar 15 '14 at 16:48
  • 1
    $\begingroup$ Also, despite being on-topic, there's nothing official in the Help page saying it is and he's looking for wording to add so we don't need to bounce people to 3 meta conversations to determine the policy. $\endgroup$ – tpg2114 Mar 15 '14 at 16:48
  • $\begingroup$ @tpg if people were flagging that question for migration then it sure did not get reflected on the comments, which were overwhelmingly on the other direction. $\endgroup$ – Emilio Pisanty Mar 15 '14 at 22:24
  • $\begingroup$ @EmilioPisanty There was some discussion in chat about it and there was at least one who indicated it should be moved. It's not that it was done unilaterally, there were people who suggested it get moved. $\endgroup$ – tpg2114 Mar 15 '14 at 22:29
  • $\begingroup$ @Nathaniel (and others) the help center actually does explicitly say computational questions are off topic. It's been that way for quite some time. So making these questions on topic does represent a change in the site's scope. $\endgroup$ – David Z Mar 16 '14 at 2:17

You must log in to answer this question.

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