While I got warnings, you are actually right that they are just warnings and it still works.

The Librarybase project is ongoing at https://librarybase.org. There is also a prototype database of citations appearing on Wikipedia at https://wikipediacitations.scatter.red. Otherwise I am not familiar with any progress on this kind of work, though it's something I am still interested in.

Demo is done; this task can be closed at any time.

The demo has been done! This task can be closed any time.

I also added a function to purge expired cache at the beginning of the script call.

Added caching for:

• the Public Suffix List
• Individual lists of domains
• Added URLs for a given combination of old revision ID and new revision ID
  • Whether a given URL matches is not cached, since that can change as the contents of the domain list changes. But whether a URL appeared in a given revision should not change (unless there was a bug in the URL extraction mechanism).

Discussion outcome: https://en.wikipedia.org/wiki/Wikipedia:Village_pump_(technical)/Archive_219#h-Gadget_proposal:_Citation_Watchlist-20250317185000

https://en.wikipedia.org/wiki/Wikipedia:Citation_Watchlist/Lists/Self-Publishing

My input and the resulting error

I now get this error:

[6aab8836cf02dc6764ce8a29] 2024-11-14 19:19:54: Fatal exception of type "Lcobucci\JWT\Signer\Key\FileCouldNotBeRead"

As part of solving this, we can store revision comparison data in browser cache. Given a combination of revisions, the URLs added are stored, and then screened against whatever copy of the list is loaded.

Still an issue as of 1.11

This should be resolved with the post-Wikimania rewrite.

I have two servers with 384 GB of RAM each, plus a workstation with 256 GB of RAM that can be used for testing. (I have tested QLever successfully on it before.) One of the servers has a bit over 2 TB free; the other has over 7 TB free. (Capacity can be added with funding.) I think either or both could be used to run QLever. What I would want to know is:

No longer occurs in latest version of Citation Watchlist.

Alternative strategy to *not* do this actually gets better results.

What I have learned, much to my delight, is that this API endpoint exists... for current revisions.

This task and T40265 should be merged since they seem to describe the same underlying problem, that if a character is treated as punctuation, it is not treated as part of the URL.

Page seems to save fine after that, though now I can't test my function because doing so requires multiline input (that's a different problem).

You can also reproduce it with this code:

import re

Also: doing the PSL and the List of Lists in one combined request

The last frontier here, I think, is batching requests for first revision ID, which is done in the event that a diff link is not available but a hist link is.

I'm not sure the database was ever in active use. In any case, I would want to rebuild it anyway, so I don't think there's any harm in deleting this particular one.

In T366815#9895547, @Pppery wrote:

But yes, this probably should be archived, as the fact that nobody noticed it was broken for years is telling.