Some time ago we had problems with Springer Encyclopedia of Mathematics, although they fixed the links after some time: Link Rot Notice: Springer EOM

Now I have noticed that many links to PlanetMath are not working. (The links which have form http://planetmath.org/encyclopedia/....)

What can we do to prevent problems like this?

  • Should we ignore the potential problems when posting and if later the links stop working, then we can update them? (This might be a lot of work if some site is frequently used in posts. But if the target of the link still exists, but with a different url, this seems to me like a cleanest solution?)
  • Or should we try to be proactive and include some kind of backup to the links we are using (for example, Wayback Machine) already when making a post? (In some cases even adding link to a Google Search might be a reasonable thing to do. If it is a reference to well-known fact, top hits among the search results might be useful for a user non familiar with it. And Google will update the links, if they change.)
  • Links and references to papers seem to be a separate issue and they have been discussed elsewhere: What is the preferred way to link references to published material?
  • 13
    $\begingroup$ Not to mention link rot due to deleted pages on this site. I have already wasted much time cleaning that up. Please be careful when deleting questions which have answers that are linked to by other pages on this site. $\endgroup$ – Bill Dubuque Jun 15 '15 at 12:51
  • 2
    $\begingroup$ This has potential of becoming a more serious problem than it is now. I have always avoided giving link answers. I'm sad to say that in my case this is to some extent also a sign of arrogance: "Only those poor sods who cannot remember (or reproduce on demand) how the argument goes resort to libraries and links." (kids, if you want to succeed in math, don't do this at home). $\endgroup$ – Jyrki Lahtonen Jun 15 '15 at 13:52
  • 6
    $\begingroup$ @JyrkiLahtonen Why do you think that closing more questions as duplicates solves the problem that I mentioned? I often link to proofs, generalizations, motivational specializations, analogies, etc. When the targets of those links are deleted then that is highly problematic, breaking crucial links of carefully constructed knowledge graphs. So now I have to avoid doing that, which does cause duplication, among other problems (nonlocalization of information, etc). $\endgroup$ – Bill Dubuque Jun 15 '15 at 14:19
  • 2
    $\begingroup$ @Bill I see your point to some extent, but really, if you're trying to craft a knowledge graph, consider using a blog or a wiki. $\endgroup$ – Lord_Farin Jun 15 '15 at 17:55
  • 8
    $\begingroup$ @Lord Eh? Are you proposing that the site can get by without links between posts? The ability to link to proofs, counterexamples, analogies, special cases provides much richer expositions than older media (such as paper books). It's your prerogative to ignore such features. But it's not your prerogative to destroy the efforts of others who use these features to construct richer expositions. $\endgroup$ – Bill Dubuque Jun 15 '15 at 18:08
  • 4
    $\begingroup$ @Bill I can see that. However, I contend that any proper answer will receive enough upvotes to mitigate the risk of deletion to an extent sufficient to safely link to it. Any answer this does not apply to is apparently not worth salvaging. However, my point was that being linked to is a factor to take into consideration when casting delete votes, however should not constitute a veto against. $\endgroup$ – Lord_Farin Jun 15 '15 at 18:24
  • 13
    $\begingroup$ @Lord_Farin: That’s not an entirely safe assumption, especially when the answer in question is at a sophisticated level in a field with a relaively small audience. $\endgroup$ – Brian M. Scott Jun 15 '15 at 19:06
  • 2
    $\begingroup$ @Jyrki: My preference when closing a question as a duplicate is to link to a previous instantiation that had one or more answers and was closed as a duplicate, if one exists. More generally, I try to maximize the number of (directly or indirectly) linked answers. I occasionally add a comment with links to very closely related questions that have good answers. $\endgroup$ – Brian M. Scott Jun 15 '15 at 19:08
  • 4
    $\begingroup$ @Lord Like close votes - it seems many deletion votes are cast without any reading of the answers. On at least a few occasions, this has caused some of my answers (to old questions) to be deleted, and some of these answers were linked to. I was able to fix some of them. But since there is no notification of deletions of old content, probably there are others I don't know about. This affects me probably more than others because I link very heavily. I think that exposing the beautiful and surprising links between various math fields is one of the best ways to spark budding mathematical minds, $\endgroup$ – Bill Dubuque Jun 15 '15 at 19:11
  • 7
    $\begingroup$ Probably it would be trivial for SE to provide some sort of notification of broken SE links when questions are deleted. Does anyone know if any such feature request has ever been made? Currently the "Linked" list is not very helpful in this regard since it doesn't distinguish links to the thread vs. links from.the thread. $\endgroup$ – Bill Dubuque Jun 15 '15 at 19:38
  • 5
    $\begingroup$ I think that @BillDubuque points out an important thing. Sometimes I also like to link other answers to enhance the quality of my own answers. E.g. : Since ... has property X it has to satisfy Y. It follows that ... and the problem is solved. The fact that property X implies Y might be a quite long and technical proof but "well-known". Now, for completeness of the post, linking a proof of the fact is, IMO, a great + to the post. Moreover, "well-known fact" seems not to be objectively well defined. $\endgroup$ – Surb Jun 15 '15 at 21:54
  • 4
    $\begingroup$ Supporting Bill's suggestion for notification of on-site link rot. $\endgroup$ – Jyrki Lahtonen Jun 16 '15 at 7:49
  • 3
    $\begingroup$ @BillDubuque There was an experimental review queue at some point that listed all broken links on the site. This was before the creation of the current review queues though, and it had some big issues. It was scrapped at some point, I'm not 100% sure why. $\endgroup$ – Mad Scientist Jun 16 '15 at 8:28
  • 4
    $\begingroup$ @quid: That any proper answer will receive enough upvotes to mitigate the risk of deletion to an extent sufficient to safely link to it, obviously. $\endgroup$ – Brian M. Scott Jun 16 '15 at 18:41
  • 2
    $\begingroup$ @quid: I did exactly that. $\endgroup$ – Brian M. Scott Jun 16 '15 at 20:54

I don't think it is necessary to provide backup links in case links eventually rot. (In case the backup links rot, do you need backup links to those, too? And so on.) But what is important is that posts should be intelligible without accessing information at third-party sources (other web-sites, books, articles, etc.).

If links in a post used only as reference rot, then it is relatively simple to fix: edit in a new link if it is easily discoverable, or remove the link with perhaps a note that "This information used to be available at ...." otherwise. In essence, in these cases link-rot is not really an issue, and can be fixed as they are discovered. But even if not fully fixed, such posts still carry useful information.

It is only a real issue when the post itself is link-only. Then it might be impossible to reconstitute it to carry information, and it should be deleted.

So the solution to link-rot, in my opinion, is to take steps to ensure that posts do not rely on information from third-party sources when they are posted:

  • leave comments explaining that posts should be self-contained, and links only used as reference.
  • if it is possible/feasible, edit the pertinent information from the source into the post itself.
  • otherwise, take steps to remove the post.
  • $\begingroup$ In case anyone is interested, the reason I give specific dates and archive information when I cite an old sci.math (or other discussion group) post of mine (e.g. "14 March 2004 sci.math post archived at Math Forum", these mostly being in comments) is because since 1999 there have been several instances in which the archives have moved --- Deja becomes google, then google-beta; Math Forum is at Swarthmore, then Drexel, then upgraded at Drexel (or maybe the upgrade occurred before the move to Drexel) --- so in anticipation of similar future changes this makes my posts much easier to track down. $\endgroup$ – Dave L. Renfro Jun 19 '15 at 18:39
  • $\begingroup$ Aurthur, surely references to fairly standard texts (let me choose Rudin to be safe) are fair game? $\endgroup$ – copper.hat Jun 20 '15 at 18:09
  • 2
    $\begingroup$ @copper.hat Depends on how it's used. "Q: How to prove Minkowski's inequality?" "A: There's a proof in Rudin's book" isn't adding much value, especially since those who have the book next to them wouldn't be looking for that question in the first place. $\endgroup$ – user147263 Jun 20 '15 at 19:42
  • $\begingroup$ @3дравыйСмысл: True, but being self contained (MSE, that is) would end up incorporating a lot of material. $\endgroup$ – copper.hat Jun 20 '15 at 19:59

You must log in to answer this question.

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