It is natural that some of the links in the posts stop working eventually. In some cases, the dead links are fixable. (And there are many editors on this site who try to edit such links.)

Perhaps we could collect here on meta various domains (that are often used in the posts on this site) where the links stopped working - but which can be corrected. We could split them into links that:

• Can only be corrected manually.
• Can be corrected in bulk by the search and replace tool.

Since Stack Exchange CMs sometimes do correct links in bulk, at least in some cases this could be used to fix the links without bumping any posts. (If we can find several domains which need correcting, it might save a bit of time for SE employees, if we ask for correcting all of them at once.) And even for the links which can only be corrected manually, having a list with a brief explanation how the links can be corrected might be useful.

Some stuff related to fixing dead links can be discussed also in the chatroom Boulevard of Broken Links.

More details on what can/cannot be done using the search-and-replace tool available to the Stack Exchange staff (which can "edit" posts without bumping):

I will add links also to some requests for mass replacements on other per-site-metas — just in case some useful information can be found also in those posts:

• A couple of instances on our own Meta regarding such broken links: (1) Link Rot Notice: Springer EOM, and (2) Old Links to PlanetMath have Broken. Apr 14 at 11:53
• I have a tangentially related idea of creating a chatroom ("Boulevard of Broken Links") to feed and triage Comments reporting broken links. Crude experiments with SEDE suggest about two dozen such Comments are posted per month. My guess is that some will be related to the domain-based issues Martin describes here, while others will have an idiosyncratic nature. Apr 14 at 16:59
• Martin, this is lovely. Thanks for all your efforts and all you do to help maintain this website. Apr 14 at 19:23
• Sigh, what happened to "cool urls don't change" ... Apr 15 at 20:15
• The chatroom Boulevard of Broken Links (mentioned in a comment above) has been created on 2022-04-15. Apr 16 at 18:10
• This can help? (didn't read the whole question above,so forgive if it doesn’t help) Apr 18 at 13:45
• Glorfindel has a Broken Link Repairer script, which is mentioned here: meta.stackexchange.com/a/375122/334566 Apr 27 at 11:12
• @PM2Ring Yes, Glorfindel mentioned this script recently also in connection with the broken projecteuclid links. Of course, this is a bit different from the bulk removal by the SE staff - edits by this script bump the posts, just as edits by a regular user. Still, it can be used for some other edits which can be automated - but which are beyond the possibilities of the SE's search-and-replace tool. Apr 27 at 13:11
• @Martin Ah, so Glorfindel is already in the loop. Excellent. :) Apr 27 at 15:56

This answer is a community wiki — do not hesitate to edit it if there are things worth adding (correcting, updating).

#### Links which can be corrected in bulk

• The links containing http://math.uga.edu/~pete do not work.

There are some exceptions, but in most cases it would be enough to change the beginning of the URL to http://alpha.math.uga.edu/~pete. (One case where it does not work that I saw was that the link http://math.uga.edu/~pete/expositions.html doesn't work after this modification, as far as I can tell, the page has been moved to http://alpha.math.uga.edu/~pete/expositions2012.html. Still, the links could be corrected in bulk — but either after the mass replacement or before, the "wrong" links should be corrected manually.)

At the moment (June 2022) I get 540 posts from search and 544 posts from SEDE. (SEDE — last activity, SEDE — already edited)

Added later: Now there is a separate post asking to replace those links in bulk — if/when that is done, this bullet point will become obsolete: Suggestion to mass-replace links to http://math.uga.edu/~pete.

• Links containing projecteuclid.org/DPubS seem to be not working anymore.

Project Euclid changed their domain structure in February 2021: Project Euclid Platform Migration FAQs. I am not sure whether this can be in some way automated. But from the link you can still get the identifier or the article and then try to use the link of the form http://projecteuclid.org/ID. For example, if the link in a post is http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.kjm/1250518067 I will use the portion euclid.kjm/1250518067 to get this link: http://projecteuclid.org/euclid.kjm/1250518067. It redirects me to the article — from there I can get often get the DOI link and/or some other links. (The FAQ post says, "All URLs on the retired platform redirect to the new platform." — but the DOI link is probably more reliable. This specific example was taken from the post Colimit of topological groups (again).)

At the moment (June 2022) I get 243 posts from search and 244 posts from SEDE (SEDE — last activity, SEDE — already edited). Probably at least some of those links could be converted automatically — we'll see what is the result of the corresponding discussion on MathOverflow Meta: Can (and should) Project Euclid links be replaced in bulk?

#### Links which can only be corrected manually

• Links to springerlink.com no longer work.

It seems that Springer has moved the domain springerlink.com to link.springer.com, but there does not appear to be any easy way to figure out where a generic springerlink.com link used to point to — unfortunately, the old links do not redirect automatically to link.springer.com.

As of May 2022, all the posts containing such links have either been edited or have a comment left under them indicating that the link is broken (along with a working replacement link when available). At this time, I get 21 posts from search and 21 posts from SEDE. (SEDE — last activity, SEDE — already edited)

• Links to front.math.ucdavis.edu no longer work.

For the links after April 2007 the correction is simple — the identifier is the same as in arXiv. The scheme used before that is more complex: "Until March 2000, the Front renumbered articles in the old mathematical archives alg-geom, funct-an, dg-ga, and q-alg as math archive articles. To avoid duplicate numbers, the system added 50 to each funct-an number, 100 for dg-ga, and 140 for q-alg." (This was found by David Roberts.) Some other details can also be found in the corresponding discussion on MathOverflow Meta. (MO had a lot of links to this site.)

As of June 2022, all the posts containing such links have either been edited or have a comment left under them indicating that the link is broken. At this time, I get 2 posts from search and 2 posts from SEDE. (SEDE — last activity, SEDE — already edited)

• Links to de.arxiv.org seem to be dead. At this time (June 2022), I get 0 posts on Mathematics and 51 results networkwide. (SEDE, SEDE — recent activity, SEDE — already edited)

• Links to eom.springer.de seem to be broken.

The problem with broken links to Springer's Encyclopedia of Mathematics was discussed earlier in Link Rot Notice: Springer EOM, however it seems that only the links to eom.springer.com were fixed at that time (or perhaps links to eom.springer.de were still working at that time).

As of June 2022, all the posts containing such links have either been edited or have a comment left under them indicating that the link is broken. At this time, I get 5 results from search and 77 results networkwide. (SEDE, SEDE — recent activity, SEDE — already edited)

• If needed, we can split this answer into several answer. (Or even add a separate section -or a separate answer - for that domains that have already been dealt with.) But since there are only three domains listed here now, this would seem a bit premature at the moment. Apr 14 at 8:26