20
$\begingroup$

In February 2021, Project Euclid made some changes on their website: Project Euclid Platform Migration FAQs.

It is probably a consequence of this that the links starting with http://projecteuclid.org/DPubS do not work anymore. (At least the ones that I have tried.) On MO I get 485 posts when I search for such URLs. If I try search in the whole Stack Exchange network, I get 950 results. (There are also links with projecteuclid.org/download - which I did not check extensively, but at least some of them work.)

To include one specific example, let us take the link http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.nmj/1118801157 which was present in the revision 1 of this answer. The link does not work - but looking at this part of the URL handle=euclid.nmj/1118801157, you can arrive at a link which actually works: https://projecteuclid.org/euclid.nmj/1118801157. (Although you can see that it was actually replaced by a DOI link in revision 2.)

Of course, it would be better to have links which actually work. At the same time, manually fixing (and bumping) about 500 posts does not seem like something which would be optimal (at least if it can be avoided). Some users have already fixed some of the links of this form here on MathOverflow - this SEDE query returns 44 results at the moment.

So the natural questions are:

  1. Can the links be converted to the short format https://projecteuclid.org/HANDLE automatically (and without bumping)?
  2. To which extent is this format of the links reliable? (If it is likely that those links stop working at some point, all work done in converting those links would be in vain.)

Concerning the reliability of the link, the FAQ post linked above says: "All URLs on the retired platform redirect to the new platform." So perhaps we could hope that links in this format should be stable.

As far as the first question goes, we should keep in mind that the tool which can be used for fixing links without bumping can only replace a specific string by another string. (Some more advanced tools - such as regular expressions - aren't available at the moment. The things which are available have been discussed a bit in another recent post on this meta.) In the example above, this would mean simply removing the substring DPubS?service=UI&version=1.0&verb=Display&handle=. So the question is whether Project Euclid links used on MO contain only a few such substrings. (Maybe it is worth looking also at other Stack Exchange sites - some of them have lot of such links, too; mainly Mathematics and Cross Validated.)

$\endgroup$
4
  • 2
    $\begingroup$ Ideally links should be supplemented with DOI links (when available) even when broken ProjectEuclid links are replaced by working ones. $\endgroup$
    – David Roberts Mod
    Apr 19, 2022 at 10:41
  • 2
    $\begingroup$ @DavidRoberts I think that there should be a FAQ posts explaining good/bad practices when citing articles. (Dead links are certainly less of a problem if the author and the title is given - as opposed to [this paper](URL).) In this specific instance, I agree that DOI links could be a good idea - bud the only way I see this being done in bulk (without bumping) is with projecuclid links. (We should keep in mind the limitations of the tools that are available.) $\endgroup$ Apr 20, 2022 at 8:08
  • 1
    $\begingroup$ Ask the site moderators to ask the Stack Exchange Community Managers to ask the company devs to do this. (Or just flag this post asking mods to add the status-review tag, and the rest will follow.) It can be done automatically, other Stack Exchange sites have successfully achieved mass link updates by asking devs to do it all in one go. $\endgroup$ Apr 20, 2022 at 13:43
  • 1
    $\begingroup$ @Randal'Thor I am aware of the link you provided - in fact I've listed it as an example here: Domains with dead links. I assume that MO users will be asked to clean up small batches first - that was the case in the previous request. But I'd hope that a CM would be willing to clean-up at least the links in the first three bullet points in bulk. (Maybe then the question we'll be whether similar thing can be done on other sites which have many links to Project Euclid.) $\endgroup$ Apr 20, 2022 at 14:10

2 Answers 2

13
$\begingroup$

Let us collect here various formats of the links to Project Euclid which no longer work. (Some of those following links also appeared in variants with https instead of http and www.projecteuclid.org instead of projecteuclid.org - as far as I can tell, such links have been edited in the meantime. Here are the searches for https and www.)

There are also some additional links which have to be edited manually. There are a few cases where something was added to the link by mistake: euclid.pjm/1102645825" in this post, euclid.jmsj/1225113599%20%22Machigashira%22 in this post.


Looking at the networkwide search, other sites which have many such links are: Mathematics, Cross Validated, Physics. Stack Overflow has a few such posts. The following sites only have a few posts with such links: Academia, Computational Science, Geographic Information Systems, Philosophy, Computer Science, Quantitative Finance, Theoretical Computer Science, Mathematica. (All edited away: Academia, Computational Science, GIS, Philosophy, Computer Science, Quantitative Finance, Theoretical Computer Science, Mathematica.)


There is also a separate problem with links to the Journal of Symbolic Logic. Many of the links to projecteuclid.org/euclid.jsl no longer work. (This is the case also for links entered directly in this format, not only the ones obtained from the transformation described above.) From the links that I tested it seems that the links containing euclid.jsl/118 are not working. Possibly this is happening because Project Euclid no longer has issues of this journal that were published before the year 2002 or after the year 2013. Search returns 20 posts on MathOverflow, 27 posts on Mathematics and 57 posts networkwide. The problematic links on MathOverflow have been edited (by David Roberts). There are some remaining links on other sites - but they at least have a comment with the name of the paper and a working link.

There seems to be the same problem with the links where the handle starts with euclid.bsl/118, again possibly because Project Euclid no longer has issues of the Bulletin of Symbolic Logic that were published before the year 2002 or after the year 2013.

The same problem also occurs in one instance of the handle euclid.jap for the Journal of Applied Probability, since Project Euclid does not have any issues published before the year 2011 or after the year 2016. (networkwide search)

$\endgroup$
2
  • $\begingroup$ Now the same search returns only the post linked in the above comment. And that particular post has comments pointing out the correct link for the Project Euclid article. So the links of this type on SO have been handled. (Thanks to the users who edited them - I only edited one such post on SO.) $\endgroup$ May 6, 2022 at 4:52
  • $\begingroup$ I think the jsl links on MO are all gone, now $\endgroup$
    – David Roberts Mod
    May 14, 2022 at 4:35
3
$\begingroup$

There's a minor bug in the tool we use to run these replacements that causes the tool to break for this particular format of link. Unfortunately, this means we won't be able to handle this particular request at this time.

We've looked at the work required to fix the bug, and it's quite possible it will be rolled into future work on the link replacement tool. That's why I've marked this : whenever we next revisit the link replacement tool, we may be able to proceed with this one. For transparency's sake, however, we don't currently have reason to revisit the tool (besides this bug), and there's no pending development work for it on our backlog, so we can't really make guarantees about when that will be.

In the meantime, you're free to choose whether you'd like to wait until the bug is addressed, or begin to fix the links now independently. (These aren't mutually exclusive options.)

Let me know if you have any questions or requests for clarification.

$\endgroup$

You must log in to answer this question.

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