Not signed in (Sign In)

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

    • CommentAuthorMariano
    • CommentTimeAug 1st 2010

    I guess Anton and Mark are reading "(last changed on YYYY-MM-DD)" as "I mean the version that came into being on YYYY-MM-DD", I tend to read "(accessed on YYYY-MM-DD)" as "This was available at least until YYYY-MM-DD", which works as some kind of ping and is somewhat useful when one is actually hunting for the reference.

    To emphasize the first reading, maybe something like "(in the version of YYYY-MM-DD)" or something would be more explicit.

    BTW, great work Anton! :)


    Anton, thanks a lot! I've already used this to add two citations to a recent paper.

    I would like to weigh in on Mariano's side concerning "last changed"; although neither am I happy with "last accessed". In fact, it seems to me that nothing guarantees that either remains true. I think that Mariano's idea of something along the lines of "(version of ...)" seems to me to be a good compromise.

    Again, many thanks for this!


    I'm with Mariano and Jose on this one. I'm not convinced by "version of ...", perhaps "version as of ..." or simply "version: ...". I do prefer having the word "version" in there as it makes it clear that there are different versions and so warns the reader up-front to go looking in the history. I've seen "accessed XYZ" to mean "The page was there on XYZ when I looked but may not be there any longer" as well as "There are different versions of this page and I mean the one that was there on XYZ".


    I agree that the word "version" is good to include. I prefer the idea of including the date on which the last edit was made, as opposed to when the citation was generated, because the point of a citation is to make it easy for the reader to find the source being cited. Presented with a list of edits and a phrase like "version of (date)" I'd like to be able to quickly unambiguously determine which version was meant, ideally by simply matching something in the citation with something on my computer screen.


    I prefer the idea of including the date on which the last edit was made, as opposed to when the citation was generated

    Certainly, but I think that this is what happens now. Yesterday I picked up two citations and the dates were those of the last edit.


    Okay, I'm convinced that the word "version" is a good idea if we're including the date the post was last modified rather than the date the post is being viewed. For now, I've changed it to "(version: YYYY-MM-DD)" since that seems clearly better than "last changed".


    Revision YYYY-MM-DD sounds better than version to me, but it's up to you!


    I've no problem with revision versus version, although now that I write that then I realise that "revision" doesn't quite work if the post hasn't been edited at all. Incidentally, I'm pleased to see the 'YYYY-MM-DD' format: I was getting all ready for a UK vs US argument ...


    Actually I've hand-tweaked the bibtex from one of the answers I got to this:

    @misc {MO32878,    
       TITLE = {Lifting units from modulus \(n\) to modulus \(mn\)},    
       AUTHOR = {Keith Conrad},
       HOWPUBLISHED = {MathOverflow},
       NOTE = {\url{} (2010-07-22)},    
       EPRINT = {},    
       URL = {\url{}},    

    which seems a little redundant in that NOTE, EPRINT and URL share most of the same information. But I suppose that different bibliography styles might ignore one or more of these fields. I use the utphys.bst, for instance, which is EPRINT-aware, but other bib styles don't seem to be.

    Could we put either a link to this thread or a summary of it in the FAQ? This question has come up on MO more than once:

    and it's likely to come up again.

    Dave Penneys suggested that citations to MO should be more like arXiv citations. That is, instead of

    Soo Key Foo (, The question title, (version: 2010-01-02)

    it should be more like

    Soo Key Foo (MO user 234), The question title, MathOverflow:1234 (v2010-01-02).

    That sounds pretty good to me. Unless there's some objection in the near future, I'll change the citation BibTeX to

    @MISC {MO:1234,    
        TITLE = {The question title},    
        AUTHOR = {Soo Key Foo (MO user 234)},    
        HOWPUBLISHED = {MathOverflow},    
        EPRINT = {MathOverflow:1234 (v2010-01-02)},    
        URL = {},    

    and the amsrefs to

        title={The question title},    
        author={Soo Key Foo (MO user 234)},    
        eprint={MathOverflow:1234 (v2010-01-02)},    
        url= {},    

    Note that this will remove the NOTE field since all that information is already in the EPRINT and URL fields.

    You can generate samples on faketestsite.


    Sounds like a good suggestion. Thanks again!


    Just tried to use one of the amsrefs citations.
    It appears that amsrefs wants the citation tag first, like this:


    @Charles: good catch. That bug has been there all along. It should be fixed now (or next time your browser replaces its cached javascript).

    • CommentAuthorTom Church
    • CommentTimeNov 11th 2010
    @Anton: I love Math Overflow, but what proportion of readers will know what to do with a reference like "MathOverflow:1234"? I've been on the site for over a year and I don't think I would understand what that means. We should remember that the vast majority of mathematicians have never heard of Math Overflow.

    I think the reference absolutely must include the URL of the page in question.

    @Tom: I can certainly see where you're coming from, but I'm not convinced (yet, at least). What is the history behind the way people cite the arXiv? When I first saw a citation like "arXiv:0811.1234v3" I was confused about how to get to the actual paper, but there's enough structure there to figure it out and there's enough culture around the arXiv that even if you don't figure it out, somebody you know knows how to dereference it. Is there reason to believe that we can't get away with the same thing?

    We can mitigate the problem by putting something in the FAQ. I could also change the 404 page so that if you try to visit it suggests "were you looking for"

    • CommentAuthorSam Nead
    • CommentTimeNov 11th 2010
    @Tom - I'm generally in favour of longer citations as opposed to shorter. On the other hand, googling "MathOverflow:30567" gives the desired page as the first hit. The same holds for arXiv pages.

    googling "MathOverflow:30567" gives the desired page as the first hit

    That's very interesting. It looks like this is far from the rule for MO, but works correctly for any arXiv paper. This is probably because the arXiv page for a paper has the literal text "arXiv:xxxx.xxxx" on it. It seems infeasible to include the text "MathOverflow:xxxxx" for every post ID which shows up in a given thread.


    (Hmmm, I thought I already replied here, but apparently not.)

    No, no, no!!! "Fake protocol" URLs like arXiv:0811.1234v and MathOverflow:1234 are evil, and should be avoided. Just because people are used to doing it for the arxiv doesn't mean we should make the problem worse. URLs should actually resolve when typed into a browser.

    In a format where hyperlinking works, it's okay if the full URL is hidden from view. Even then, it would be better to not display a string with a colon in it, just so that people don't start pretending it's a proper identifier.


    @Scott: How about stripping the http:// and the URL from the user ID? Two URLs in a single citation is annoying and http:// is an eyesore.

    Soo Key Foo (MO user 234), The question title, (version 2010-01-02).


    I don't much care about the user number; I'm happy whatever we do there.

    I have a slight preference for keeping the http:// for the question URL; I mind it less, and it's a clear marker that the string you're looking at is actually a URL.

    • CommentAuthorJDH
    • CommentTimeNov 13th 2010
    The cite button doesn't appear for me under Windows Internet Explorer 8, although it does appear on the same machine using Firefox. I thought I recall it working fine under IE8 earlier, so perhaps some recent change has upset things somehow?

    Many of the public machines that I use to access MO only have IE8 and no possibility to install another browser.

    What's the point of including the question number instead of the URL? That doesn't help anyone. Mathoverflow questions are essentially always Google-able by the title, so either just have that, or give the full URL.


    Since I used some MO citations recently, I decided to tweak them slightly. They now look like this.

    @MISC {MO1234,    
        TITLE = {The question title},    
        AUTHOR = {Soo Key Foo\phantom{x}(},    
        HOWPUBLISHED = {MathOverflow},    
        NOTE = {\url{} (version: 2010-01-02)},    
        EPRINT = {},    
        URL = {},    

    The main changes are (1) there was no \url before, so you didn't get a link in your pdf, and (2) there's now a strange \phantom{x} whose purpose is to prevent BibTeX from thinking that the person's last name is ( and therefore generate a stupid key like [(ma] instead of [Foo]. My reasoning is that even though this is crufty, it will produce the desired output more often (most papers I read use the alpha bibliography style). If I'm off base, somebody let me know and I'll change it back.


    Anton: Funny you should mention this, because just now I had this problem with a citation to StackOverflow in a paper I am writing and which used the amsalpha bibliography style. I was wondering why it generated the stupid key. My solution was to switch to "utphys" which is my default bib style anyway.

    Anyway, I think that the changes are good. I had already added the \url myself, so this saves me having to do it by hand.

    Cheers, José

    PS: and congratulations on completing your dissertation!