Not signed in (Sign In)

Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.

    This question:

    was closed with the comment "Sorry, but questions on specific software (even if mathematical) are not on-topic on this site." I have no desire to reopen that particular question, but the general policy has been mentioned in connection with other questions and I am unclear about its status. To me it seems arbitrary and unhelpful. It is not mentioned in the faq and I could not find any previous discussion of it on the meta site.
    Neil -- whether or not questions about specific software are appropriate for this site, for pretty much every piece of software I've ever used there was a place on the internet that was *perfect* for questions about it (in the sense that the place was specifically designed for this purpose), and in no cases was it this site :-)
    • CommentAuthorquid
    • CommentTimeAug 21st 2011 edited

    Since I wrote the comment in question, let me elaborate. The original comment is terse, and perhaps it was not phrased well.

    What I meant to say, in accordance with Kevin Buzzard, is merely (and I believe there are at least precedences for this) that MO is not to be used as a substitute for a support-mailinglist for whatever (math) software. Indeed, I thought about writing instead "MO is not a support-mailinglist." But this felt somehow unfriendly, and so I wrote that instead.

    Now, it is true, that I can imagine questions of which one could say they are on a specific math software that I would consider as on-topic on MO. So the precise wording of the comment is not good.

    But, what I think is essentially always off-topic is a question of the form: I entered [this piece of code] in software X. Why does this do A (rather than B)?

    Later Addition: Rereading it seems to me I did not really address the actual question, that is why I consider such questions off-topic according to the written guidelines of MO.

    First, in most cases, such question are not a research level math question (and thus by the FAQs are already outside the primary focus of MO), and they are also (in most cases) not something of the form which while not being a rlmq is still a good MO question. For example, as they tend to be very localized. (What to me for example could make a good MO question, and questions like this exists, are of the form: I am trying to solve computational problem P / am working in field F. Which software would be best suited for me / there are various different tools I known of what are pros and cons? [However, this then already is not a question on a specific software, and thus would not fall under what I wrote. And, this was also a motivation for writing the 'specific' in the first place.])

    Second, there is a request in the 'how to ask page' along the lines of 'doing ones homework' before asking on MO. The question of the form I described at the end of my original post, should almost always fall short of this, as typically there are more or less obvious, and not too much more inconvenient, other places to get an answer. (In particular, if a questioner would indicate somehow why s/he is using MO for the question, the situation would be different. For example, I in principle could imagine a situtation where somebody wants to use a highly speciliazed piece of OSS math software and the original developper is unreachable for some reasons and there actually is no default place to get an answer.)

    So, in summary, I think that there are various reasons why the typical question on a specific software is off-topic. And, thus why I think in general such questions are off-topic. (Though, I could imagine unusual circumstances/exceptions.)

    In case nobody objects or comes up with a different point of view, I will in the future add an 'in general' to comments of the form mentioned, but otherwise follow the policy I tried to sketch above, which is the one of which I thought (perhaps wrongly) was generally accepted. (Needless to say, in case this discussion should show that this was a misconception or there is a desire to change this policy, I will follow another policy.)


    Interestingly, math.stackexchange DOES allow questions about software that mathematicians use. I remember a recent one about "why does Mathematica not simplify $0^x$ to $0$?" or something.

    • CommentAuthorquid
    • CommentTimeAug 22nd 2011

    Gerald Edgar, I did not look for the specific question you mentioned, so go purely from your description: to me this question is, interepreted in its spirit, not a question of the form we are discussing. Since to answer it in a meaningful way, one has to interpret it anyway as "What are reasons why a CAS would/should not simplify 0^x to 0 ?" or something along these lines. Because otherwise how would you want to answer the question except by saying: "Becasue the person(s) writing the relevant code decided so."

    Or in other words, it is not really about a specific software, but more general; Mathematica could be replaced by (certain) other CAS, and the question would remain meaningful.

    • CommentAuthorWillieWong
    • CommentTimeAug 22nd 2011

    Interestingly, math.stackexchange DOES allow questions about software that mathematicians use.

    I do tend to post a comment suggesting to the users the appropriate forum or mailing-list, especially for big-budget software like Mathematica, for the reason that they are likely to get more prompt and better help there than from the Math.SE community.

    Ideally we should have some statement about software questions in our FAQ. However, I've been under the impression that "tools of the trade" questions are okay, especially as CW. For example, of all the questions I've asked on MO, this one received the most votes. So a proper phrasing in the FAQ will be a bit delicate.

    Consider this question, for example:

    There are two votes to close, and Andrew has left a comment suggesting that it should be taken to a Matlab-specific site. I disagree with this. Without actually knowing the answer to the question, it is hard to be sure about it. It is possible that there is a resolution that is well-known to Matlab experts, but it seems equally possible to me that there are serious mathematical issues involved which might be better addressed on MO. I grant that MO's expertise in such areas may be relatively thin compared to set theory or algebraic topology, but I see that as an unfortunate historical accident rather than something we should be reinforcing.

    FYI, I do have a MO question about software:

    Technically, it's about the algorithms that SnapPea use so it's not a strict usage question.

    • CommentAuthorquid
    • CommentTimeAug 22nd 2011 edited

    I think the main problem with this new question (73392) is that it is very vague, as noted in the answer by fabee. Perhaps there is an interesting math problem, but how should one know, let alone help to solve it from the information that is given? This seems plainly impossible. Now, perhaps the questioner will add more details and then it might be a good question. But, at the moment this seems way to vague for me, whether on software or not.

    Still, I did not (yet) vote to close this one, mainly due to the "Or, can you give me some suggestions?" [making it a general question] and the fact that the questioner might clarify in view of fabee's answer.

    And, regarding Andrew's comment: the questioner seems to mainly ask for a replacement/addition of one matlab tool by another matlab tool. I used matlab the last time more than a decade ago and never used it seriously, but still I am quite sure, and a quick google search seems to confirm this, that there are plenty of places on the web were nothing but matlab is discussed. So, it seems simply wiser to ask there. So, Andrew's comment seems like good advice.

    Finally (this is not meant as critical as it might sound): to some extent I find this discussion difficult, because for both questions there is at least one reason other than it being about a specific software that in my opinion definitely would warrant a closure. Is there any good or at least reasonable question that got closed just because it was about specific software?

    Added after seeing Ryan's comment: to me this cetainly seems like a good question (and it seems also nobody wanted to close it). But, as Ryan says, it is not a usage question, but in spirit a question on an algorithm. And, say, if somebody asks an advanced question on symbolic integeration and frames the questions along the lines the lines of the implemantion of some algorithm in some program, then I think this person will not be faced with much problems either. Since such questions are not only about using a specific software.


    For me, it's not just about "Is this question suitable for MO?" but also "Is MO the best place to get an answer to this question?". In general, for software, I really feel that asking a bunch of mathematicians is not as good a place to get an answer as asking a bunch of programmers; especially if those programmers are involved in the program itself.

    Thus for a software question I'd want to see some evidence that MO really was the right place, that there was something special that only a mathematician could be expected to know, and not just a convenient place. In particular, to lift a sentence from Neil's last comment:

    Without actually knowing the answer to the question, it is hard to be sure about it.

    For a question that isn't obviously on-topic, the onus on the questioner is to make it obvious to the average MO user that it is on-topic for this site. So this sentence of Neil's sums up why I voted to close that particular question and suggested an alternative avenue for an answer.

    • CommentAuthorNilima
    • CommentTimeAug 22nd 2011
    I agree, MO is not the best place, in general, to get answers to software questions.

    Here's a question I asked on a software forum, with no response:

    I'd like to generate sample statistics and trends whilst varying a parameter, for the behavior of solutions of a nonlinear parabolic PDE with additive white noise. We conjecture that as the mesh is refined, the solution fails to converge in any reasonable space. We'd like to test this conjecture. I'm currently using a pseudospectral discretization in space with an Euler-Marayama time-stepping scheme. In order to gather meaningful statistics (the trend I conjecture is weak), I need on the order of 1000 samples, which I intend to gather by running code on different machines on the network. Clearly tying the seed to the clock will not be a good solution. Any pointers on which of the Matlab random number generators would be most robust?

    Someone from the site wrote saying my question was very academic and had not arisen in that form before. I tend to agree, the question is less about the software than the best way to use it for the purpose of testing a conjecture.

    However, I wouldn't ask this question on MO.

    The average MO user is not a numerical analyst, and may not be able to discern if this is on topic. There would likely be a lot of back-and-forth about this. Arguably the only MO users whose input is relevant is that of fellow numerical analysts/computational scientists, and of these there are few. It is currently more efficient to ask a colleague about such things.
    • CommentAuthorWill Jagy
    • CommentTimeAug 22nd 2011
    Hi Nilima. Let me draw your attention to

    where I did not vote to close (nobody did) on the theory that there might be something to it.
    • CommentAuthorNilima
    • CommentTimeAug 22nd 2011
    Hi Will, I think it's good that quid helped clean up the question, and that you pointed the OP to SE. The question 73371 is not about software, though.
    • CommentAuthorWill Jagy
    • CommentTimeAug 22nd 2011
    Agreed, 73371 is not about software. Originally I put a link to the Wikipedia page about the MP pseudoinverse, but another MO user, well, mocked my comment, so I deleted that. Hard to know what else to say to 73371, though.

    Dear Nilima,

    There has been a lot of talk on how to attract qualified people from different fields to MO. The only consensus so far is that we need good questions in these fields in order to attract researchers to answer them. While what you describe in your last paragraph sounds reasonable at first sight, a better approach would be to post the question on MO and then email your colleagues with a link to the MO question. That colleague will become aware that MO is a place where you can ask numerical analysis questions and will hopefully consider joining the MO community, adding one more numerical analyst to the flock. There are side benefits too, such as the fact that MO is much better at rendering math than plain email.

    • CommentAuthorNilima
    • CommentTimeAug 22nd 2011
    Dear Francois,

    I personally got burnt with this approach recently. I found an intriguing question on MO, and emailed a non-MO colleague a link to it. By the time the colleague looked at the question, a bit of turbulence had appeared in the comments. Needless to say, said colleague was not enamoured of what was on display (neither was I).

    However, this constitutes a single anecdotal data point, and does not detract from the sensible principle you suggest.
    • CommentAuthorWill Jagy
    • CommentTimeAug 22nd 2011 edited
    Never MInd
    • CommentAuthorquid
    • CommentTimeAug 22nd 2011

    In view of the discussion, in particular Nilima's example, I will be more careful with statements along the lines 'software not on-topic'; since indeed it is more general than what I mean to say, and (thinking about it) I can see how it can be harmful to state this overgeneralization (even if the immediately affected question itself is one that should be closed). [Thus, I also deleted the recent comment I made along these lines.]

    • CommentAuthorNilima
    • CommentTimeAug 22nd 2011
    Dear Quid, I concur that questions about, say, syntax or 'can you debug this code' are not appropriate. There are also more mathematical questions about software (about computational complexity of a certain variant of an algorithm, say). Distinguishing between these may prove tricky unless there's a critical mass of relevant expertise. Such expertise in other sub-disciplines allows community moderation to be effective.

    I actually believe that at least for now (ie, August 2011) the strategy to *not* allow software-related questions is reasonable. Interested and energetic numerical analysts/scientific computing people will have other questions to attract them to MO. It may be unwise to have a lot of low-end 'help me debug' questions with relevant tags floating around.

    I'm sure such questions will arise periodically, and the issue can be revisited at a later time. In the meanwhile, maybe an informal 'let it sit for a N hours to see who bites, else close' approach may be useful. On the specific question under consideration, user fabee provided a reasonable answer.
    • CommentAuthorquid
    • CommentTimeAug 23rd 2011

    Dear Nilima, thank you for this information. It is very valuable for me.


    Regarding the question It strikes me that this is really a question about algorithms, which is why it feels different from the usual software questions. I think the question would be better if the OP asked for an appropriate algorithm rather than an implementation thereof.