Wikisource:Administrators' noticeboard

Administrators' noticeboard
This is a discussion page for coordinating and discussing administrative tasks on Wikisource. Although its target audience is administrators, any user is welcome to leave a message or join the discussion here. This is also the place to report vandalism or request an administrator's help.
  • Please make your comments concise. Editors and administrators are less likely to pay attention to long diatribes.
  • This is not the place for general discussion. For that, see the community discussion page.
  • Administrators please use template {{closed}} to identify completed discussions that can be archived
Report abuse of editing privileges: Admin noticeboard | Open proxies
Wikisource snapshot

No. of pages = 2,943,333
No. of articles = 775,440
No. of files = 22,642
No. of edits = 9,959,864


No. of pages in Main = 461,300
No. of pages in Page: = 2,104,374
No. validated in Page: = 421,285
No. proofread in Page: = 689,966
No. not proofread in Page: = 803,349
No. problematic in Page: = 29,715
No. of validated works = 3,478
No. of proofread only works = 2,585
No. of pages in Main
with transclusions = 253,740
% transcluded pages in Main = 55.01
Σ pages in Main


No. of users = 2,934,620
No. of active users = 348
No. of group:autopatrolled = 463
No. in group:sysop = 27
No. in group:bureaucrat = 2
No. in group:bot = 22


Checkuser requestsEdit

  • Wikisource:checkuser policy
  • At this point of time, English Wikisource has no checkusers and requests need to undertaken by stewards
    • it would be expected that requests on authentic users would be discussed on this wiki prior to progressing to stewards
    • requests by administrators for identification and blocking of IP ranges to manage spambots and longer term nuisance-only editing can be progressed directly to the stewards
    • requests for checkuser

Bureaucrat requestsEdit

Page (un)protection requestsEdit

Request protection of Main Page templatesEdit

According to the very first point under Wikisource:Protection_policy#Special_cases “The main page should always be protected…”, yet this edit took place today. Some care and attention, please? (Normally Phe-bot is the sole updater of Template:ALL TEXTS!) 114.73.248.245 17:57, 15 October 2018 (UTC)

Thanks. I have place soft protection on the page. — billinghurst sDrewth 20:09, 15 October 2018 (UTC)
  Comment To fellow administrators, I have up'd the protection on a couple of templates that won't need updating. I have a question about Template:Highlights, should this be sitting at semi/soft? If we are unlikely to change it, then we should be protecting it further. — billinghurst sDrewth 20:27, 15 October 2018 (UTC)

OtherEdit

Resource Loader issue needs outside guidanceEdit

The more I read up on this RL change and the subsequent actions needed (or taken?), the more I get the feeling some of my approach to site wide & gadget .js/.css organization over the months is going to behind this week's latest problems. If that winds up to be the case, then I'm truly, truly sorry for that. Let me try to document those steps and the reasoning behind them in hopes someone (@Krinkle:) can made sense of our current state and put us on the right path post RL change(s).

Originally, we not only had a ridiculous amount of scripting and .css definitions in our primary site-wide MediaWiki files to begin with but also called a number of stand-alone .js/.css files within those primary MediaWiki files called unnecessarily in addition to calls to various sub-scripts on top of any User: selected gadgets being called -- some of which eventually became default loaded per concensus, etc..

A simple depiction of the key files mentioned minus any Gadgets basically went like this...

Over several months with help of other folks, I began to consolidate and/or eliminate as much scripting calls as I could -- creating optional Gadgets whenever possible -- and tried much the same for the .css class definitions. The rationale behind doing this can be found in several places, most importantly: Wikipedia. The premise to keep the MediaWiki site-wide files "lean" goes like this....

 /**
 * Keep code in MediaWiki:Common.js to a minimum as it is unconditionally
 * loaded for all users on every wiki page. If possible create a gadget that is
 * enabled by default instead of adding it here (since gadgets are fully
 * optimized ResourceLoader modules with possibility to add dependencies etc.)
 *
 * Since Common.js isn't a gadget, there is no place to declare its
 * dependencies, so we have to lazy load them with mw.loader.using on demand and
 * then execute the rest in the callback. In most cases these dependencies will
 * be loaded (or loading) already and the callback will not be delayed. In case a
 * dependency hasn't arrived yet it'll make sure those are loaded before this.
 */

The result of that effort as it stands today can be depicted basically like this....

The predominant change in order to move towards the previously cited rationale & approach is that the bulk of the scripting and class definitions now reside in the default-enabled Site gadget files, MediaWiki:Gadget-Site.js & MediaWiki:Gadget-Site.css. And by no means is the current state the desired final approach; its been a work in progress as time allowed over several months.

Obviously, now with the recent change to Gadgets and ResourceLoader, either the existing rationale or my attempts (or both) are no longer in harmony -- if they ever were. In my view, we need someone like Krinkle (or maybe the collective minds of Wikitech-l?) to take the time and attention needed to come in here and straighten all this out -- one way or the other. My gut tells me THAT will resolve the reported loss of one thing or another post-RL change(s). Again, if I'm right about my actions exacerbating problems for other, I apologize and take full responsibility. -- George Orwell III (talk) 20:54, 8 August 2015 (UTC)

I've made a few minor changes in addition to yours that hopefully make things work a bit more like you intended. I'm happy to provide further guidance but that probably works better for a more specific need or question. Perhaps bring it up on Wikitech-l or on IRC so we I can help you move forward with any unresolved issues. Krinkle (talk) 21:37, 20 August 2015 (UTC)

Interface administratorsEdit

Hi. Please see https://www.mediawiki.org/wiki/Topic:Unisfu5m161hs4zl. I do not remember if this was already discussed and how it is going to be addressed. Comments and suggestions welcome.   Comment As far as I am concerned I would trust any admin who feels skilled and confident enough to tackle such edits.— Mpaa (talk) 21:05, 29 October 2018 (UTC)

I can handle the technical aspects of it. However, it can take me a while to get around to tasks that take longer than a few minutes, so I don't want to create a false expectation of being able to handle time sensitive matters on my own. —Beleg Tâl (talk) 02:35, 30 October 2018 (UTC)


We should decide how to address the fact that EnWS has no m:interface administrators. I see basically the following options. Please add/amend as you feel appropriate.

Option A - Assign right on demand when needed

Option B - Assign right permanently to willing Admins, to be reviewed in the confirmation process

As I said above, I am for the simplest one.— Mpaa (talk) 21:28, 30 October 2018 (UTC)

Option C - Assign right permanently to selected Admins, after approval process, to be reviewed in the confirmation process

Option C sounds like you're being volunteered (based on the lack of the word 'willing'). ;) --Mukkakukaku (talk) 06:27, 31 October 2018 (UTC)

Option D - assign the rights to all the admins, who have already been vetted for community approval, and then whoever has the ability and desire can make use of it as they will and as needed. —Beleg Tâl (talk) 13:33, 31 October 2018 (UTC)

Option D would make the most sense for us. For anyone to get themselves to the point that we trust them with the admin tools just so that they can mess around in the interface, they would be playing a very long game. Beeswaxcandle (talk) 22:05, 2 November 2018 (UTC)
I agree with Beeswaxcandle, Option D, although I would also be fine with the right only going to admins who express an interest. BD2412 T 23:00, 2 November 2018 (UTC)
It is so rare I disagree with Beeswaxcandle but this must be one of those times. The whole point of this change is to prevent the ignorant from accidentally screwing up - insulting as the implications undoubtedly are! As such under the new regime trust is no longer enough; perhaps somebody ought to draw up some kind of eligibility examination…? 114.73.248.245 23:03, 2 November 2018 (UTC)
That hasn't been an issue for us yet, and accidental changes are easily reversed. If we had more users it would be more of a problem, but as it stands this kind of distinction is more cumbersome than helpful in my opinion. —Beleg Tâl (talk) 00:08, 3 November 2018 (UTC)
As much as I like the idea of making all existing admin interface admin, IA were separated from regular adminship specifically to reduce attack surface(from hackers), and it was pretty dangerous if the access fell into the wrong hand, I'd rather propose having existing admin request right from bureaucrat and could be granted at the bureaucrat's discretion, and should be automatically removed if no action after two month.Viztor (talk) 02:13, 10 August 2019 (UTC)
  •   Comment we discussed it when the rights were split, and it was agreed that it could be assigned on a needs basis. That has been done at least once for me with the temporary assignation of the IA rights. — billinghurst sDrewth 05:58, 10 August 2019 (UTC)
    Note that WMF Legal requires 2FA to be enabled for users who are to be assigned this right, so bureaucrats will have to verify this before doing so. MediaWiki's 2FA implementation is also sufficiently finicky that one may not want to enable it without proper consideration. --Xover (talk) 08:21, 10 August 2019 (UTC)
    What's wrong with the 2FA implementation? I haven't had any issues with it at all. —Beleg Tâl (talk) 22:17, 10 August 2019 (UTC)
    Ah, sorry, I should have been more clear. I am going on hearsay, mostly from admins on enwp (a crotchety bunch if ever there was one), and my own assessment of the documentation at meta. The main complaints are that the implementation in general is a little bit primitive (as is to be expected since WMF rolled their own instead of federating with one of the big providers), and that there is no way to regain access to your account if something goes wrong with the 2FA stuff (if your phone is stolen etc.) unless you happen to know one of the developers personally. None of these are in themselves showstoppers, and many people are using it entirely without issue. The phrasing sufficiently finicky that one may not want to enable it without proper consideration was not intended to discourage use, but merely to suggest that it is worthwhile actually giving it a little thought before requesting it be turned on. --Xover (talk) 17:52, 11 August 2019 (UTC)
    Okay, gotcha. As it happens, Wikimedia 2FA does include emergency access codes for use when your phone is unavailable. —Beleg Tâl (talk) 19:56, 11 August 2019 (UTC)

Formal requirements related to 2FAEdit

Picking up this again…

I finally got so annoyed by our inability to fix even simple stuff stuff that requires Interface Admin permissions that I hopped over to meta to figure out what the actual requirements are (versus the should stuff). As it turns out, the 2FA stuff is (surprise surprise) as half-baked as most such Papal bulls from the WMF: 2FA is required for intadmin, but there is no way for bureaucrats to actually check whether an account has that enabled. The result of this is that even on enwp (where they take this stuff really seriously) they do not actually try to verify that 2FA is enabled before they hand the permission out: they check that the user is in the right group so that they can turn on 2FA, remind the person in question of the requirement, but otherwise take it on faith (trust). There's a request in for the technical capability to verify 2FA (and I think Danny is even working on it), but it seems mostly everyone's waiting for 2FA to be enforced by the software.

Meanwhile, anyone with existing advanced permissions (i.e. +sysop) have the capability to enable 2FA, and anyone with a particular reason (e.g. that they need it to get Interface Administrator permission) can apply to be a "2FA Tester" and thus gain the ability to turn it on.

The net result is that our bureaucrats (ping Hesperian and Mpaa) can assign this permission so long as we somehow somewhere make at least a token effort to make sure those getting the bit have 2FA enabled. Whether that's an addition to, or footnote on, Wikisource:Adminship, or the bureaucrats asking/reminding the user when it comes up, or… whatever… I have no particular opinion on. Since the previous community discussions have been actively adverse to regulating this stuff in detail, and absent objections, I think "Whatever Hesperian and Mpaa agree on" is a reasonable enough summary of consensus.

I still think we should have an actual policy for Interface Administrators (or section on it in Wikisource:Adminship) and some facility for permanently assigning the permission (ala. +sysop; but intadmin tasks are not one-and-done like +sysop tasks, they often require iterative changes over time and need to fit into a overall architecture), but so long as there is no appetite for that, something that we can point to and say "That's how we handle the 2FA requirement" if the WMF should ever come asking. --Xover (talk) 07:37, 10 February 2020 (UTC)

Judicious cleaning required from Special:UnusedFilesEdit

I was just poking my head into Special:UnusedFiles. There are a significant number of images that utilise {{raw page scan}} that should be checked and if truly unused, we can delete as the file has been transwiki'd to Commons. And I note that always physically check their usage as I have previously seen that the NOT USED assessment is not always accurate.

Checking and deleting process:

  • Use the Page:… link
  • At Page:… check that there is a Commons loaded image in place (and no use of template:raw image)
  • grab the new filename
  • click back to the local image, delete the image, noting "File transwiki'd" and paste in the new filename (preferred not mandatory)

If admins could do 10 to 20 a session, we should get through them in a month or so. — billinghurst sDrewth 09:26, 14 November 2019 (UTC)

  Comment TIP: when doing the image check you can even take the time to validate proofread page with the image (very often sittin gin proofread status). — billinghurst sDrewth 09:31, 14 November 2019 (UTC)
@Billinghurst: Considering the list is maxed out at 5k, so there's no telling how many more of them there are (we have over 20k files tagged as raw page scans, possibly more that are untagged), it's highly unlikely we'll get through that in a month. But it's certainly something we need to start chipping away at.
And we should possibly even start considering more drastic measures, like periodically bot-deleting anything in Category:Raw page scans for missing images that isn't used anywhere (including inbound links). The manual processing is tedious and time-consuming, and provides very little additional value compared to an automated approach (linking to the replacement image in the deletion log, mostly, and that has marginal value at best). We'd need to check closely whether the category contains files that could be caught as false positives in such a run, but barring such pitfalls automation may be both the best option and the only realistic way to ever clear out this backlog (we have plenty of other image-related backlogs where human attention is necessary).
Oh, PS, DannyS712 has a neat user script at User:DannyS712/Change status.js that makes cases such as this a lot quicker. I'm not sure they consider it ready for prime-time (I don't think it's been advertised anywhere), so caveat emptor, but I've been using it a good bit today and seen no problems. To use, add importScript('User:DannyS712/Change status.js'); to your common.js. --Xover (talk) 20:18, 14 November 2019 (UTC)
If you want to use the script, it adds a link to the function (next to the "move" function) that will, if the page is "not proofread" or "problematic", mark it as "proofread". If it is already "proofread", and the user can mark it as validated, it marks it as validated instead. Let me know if there are any questions. Thanks, --DannyS712 (talk) 20:34, 14 November 2019 (UTC)
@Billinghurst: I agree with @Xover: that this kind of tasks should be automated.Mpaa (talk) 21:29, 15 November 2019 (UTC)
  Comment At this point I would think that the task is to start to chip away. I don't see that there is urgency in cleaning this space, so as long as we start. So what if it takes three months, heck I have works that I dip in and out of for years. As I said I have seen multiple issues of the tool being wrong in the past, if we can demonstrate that this is no longer the issue, then maybe we can look to bot removal. I though the admin review, and process of validating was beneficial.

P.S. Those quiescent admins, and those who find it hard to identify tasks to undertake are given a gift here! — billinghurst sDrewth 22:50, 14 November 2019 (UTC)

Having done about a hundred of these by hand… I'd say the realistic best case sustained rate here is something like 5 admins doing 5 files per day for 5 days a week. That's an aggregate rate of about 500 per month. If the number is 5k that means 10 months to get through it. If the number is 20k that's 40 months, or just shy of 3.5 years. I don't have sufficient data for an accurate estimate of net time, but assuming a range of 30-60 seconds per file, at 5k files that's an aggregate ~40–80 net admin-hours expended. At 20k files that's ~160–320 net admin hours. Assuming an 8 hour work day, that's one dedicated admin working flat out only on this for between one week (5k/30s) and 2 months (20k/60s). With no lunch break, by the way. That's a pretty high cost.
On the upside we have tagging the logs for deleted files with a link to the replacement images. But that only matters if you're actually looking at the deleted file, and for these raw page scans that is essentially never going to happen. Having a human in the loop also helps guard against Mediawiki bugs in categorising etc., but while, yes, that does happen, it's been years since I've run into that kind of bug anywhere that would matter here. What usually happens is that counts and references fail to update properly when pages are deleted, so you get categories saying they have members, but in reality the relevant items have already been deleted; and these eventually get cleared out by periodic maintenance tasks.
In other words, doing this manually is expensive and with a significant opportunity cost, and without a concomitant value. Automating it obviously carries risks (automatically deleting up to 5-20k files should never be done lightly). But with appropriate checks—for example, all files listed on Special:UnusedFiles who are also in category Raw page scans and who have no incoming links in WhatLinksHere—manual spot checks, and going in batches… the risk should be eminently manageable. --Xover (talk) 09:16, 16 November 2019 (UTC)
I can work out shortly a script that can scan category Raw page scans and checks for the conditions for deletion (and in case deletes). If you are OK to test small batches, let me know.Mpaa (talk) 14:52, 16 November 2019 (UTC)
@Mpaa: Can you have it run up a list of files and dump it in a sandbox somewhere so we can spot check the logic? Maybe a hundred or so files that the script thinks should be deleted, and, if relevant, the ones it thinks shouldn't be. Better to find any holes in the logic before we start deleting stuff. --Xover (talk) 15:10, 16 November 2019 (UTC)
Here: res sandbox.Mpaa (talk) 16:36, 16 November 2019 (UTC)
@Mpaa: Excellent! I've spot-checked pages from most of the works represented in that list and found none incorrect. I'd have no objection to running that (in batches so it can be checked; there're bound to be some pathological edge-case out there somewhere). --Xover (talk) 17:12, 16 November 2019 (UTC)
@Xover: I have done a small test batch of 45 pages as Mpaa.Mpaa (talk) 18:29, 17 November 2019 (UTC)
@Mpaa: Ok, I've spot-checked 2–3 files from each work in those 45, and find no real problems. The only issue I see is that the deletion log for File:A book of the west; being an introduction to Devon and Cornwall.djvu-453.png links to c:File:A book of the west; being an introduction to Devon and Cornwall.djvu instead of c:File:A Book of the West - ALMS HOUSES, S GERMANS.png; and ditto for File:A book of the west; being an introduction to Devon and Cornwall.djvu-223.png that points to c:File:A book of the west; being an introduction to Devon and Cornwall.djvu instead of c:File:A Book of the West - LAKEHEAD, KISTVAEN.png. --Xover (talk) 21:28, 17 November 2019 (UTC)
@Xover: Thanks, I fixed it, I ran another ~40 pages.Mpaa (talk) 18:33, 18 November 2019 (UTC)
I'm wondering if we can set up a scrolling gallery that does nothing but compare our page image side-by-side with the comparable Commons file. An editor could scroll through and eyeball any differences fairly quickly. BD2412 T 22:52, 18 November 2019 (UTC)

┌───────────────────────────────────────┘
Poking at this again…

I found no problems with Mpaa's test bot run, and we still have potentially ~20k files sitting there that it would be a waste of admin resources to process manually. Can we pull the trigger on a mass delete of these? If not, what are the concerns? --Xover (talk) 08:26, 18 February 2020 (UTC)

I ran about 15 pages to check that everything is still OK. Mpaa (talk) 21:47, 18 February 2020 (UTC)

Title blacklist updated to prevent invisible characters in page namesEdit

Based on this discussion I have added rules to the title blacklist to prevent the creation of pages with invisible Unicode characters in the name. Users hitting this rule should see the custom error message at MediaWiki:titleblacklist-invisible-characters-edit. --Xover (talk) 09:05, 5 December 2019 (UTC)

Should the “Emojis, etc. Very few characters outside the Basic Multilingual Plane are useful in titles” section be there? First, it’s clear that MediaWiki:titleblacklist-invisible-characters-edit is not an helpful message in that case. Secondly, there seems like some quite useful characters there; the Mathematical Alphanumeric Symbols is basically just so we can include mathematical titles in a plain text format. As for the rest of them, if we allow Chinese, we should allow all of Chinese; if we do enough academic work, we’re going to have Hieroglyphics and ancient Chinese in article titles.--Prosfilaes (talk) 11:26, 5 December 2019 (UTC)
@Prosfilaes: In page names? Where we recently had community consensus to not even permit curly quotes (as part of the discussion to permit them in page content) because that’d be too fiddly? In any case, both the error message and the rules can be tweaked if needed. The current rules are an attempt to prevent stuff like WORD JOINER and friends that are invisible and cause issues for people trying to work with such pages. I (currently) have no particularly strong opinions on the issue above a vague inclination towards limiting page names to roughly ASCII (on enWS, of course, other language projects have different needs). --Xover (talk) 11:52, 5 December 2019 (UTC)
What’s magical about page names? Page names need to match the names of the works they contain. Curly quotes are a special case; note that there was no restriction on character in pages, except for curly quotes. If an article title is "Čapek’s works in English” or “Injections from 𝕎 to 𝕁", then why should the page name be any different?--Prosfilaes (talk) 13:55, 5 December 2019 (UTC)
Page names have technical and practical concerns (peoples' ability to enter them, display, them, search for them, etc.) that means we should constrain them in some fashions; and at the same time we already do stuff like drop “The” from page titles to facilitate automatic sorting (which I actually disagree with, but that’s neither here nor there). The characters we disallow in the current rules are also exceedingly rare in practice, and the blacklist can be overridden by any admin at need, so I don’t think it is a problem we should expend too much effort on until and unless we start seeing actual cases where it causes problems.
Č (U+010D: LATIN SMALL LETTER C WITH CARON), is in the Latin Extended-A block which is a part of the Basic Multilingual Plane which the current ruleset lets through. 𝕎 (U+1D54E: MATHEMATICAL DOUBLE-STRUCK CAPITAL W) and 𝕁 (U+1D541: … J) are part of the Mathematical Alphanumeric Symbols block of the Supplementary Multilingual Plane (most common mathematical symbols are in the BMP; these are essentially font variations: bold, italic, fractur, etc.) that contains all the exotic and ancient stuff (Linear B, Coptic, Hieroglyphs, etc.) plus a good chunk of Emoji, so they would be disallowed by the current rules but we can whitelist ranges if we discover that we need them.
That all being said, I am by no means married to the current rules so whatever is the consensus is is fine by me; and I would, of course, be happy to help implement whatever that consensus is if needed. Most of the examples you mention above (the extended maths stuff, hieroglyphics, etc.) are contained in distinct blocks (Emoji aren’t, and I believe Chinese is also split up in inconvenient ways when you want to handle everything, but most of the rest) that should be relatively straightforward to whitelist. --Xover (talk) 13:32, 6 December 2019 (UTC)
I agree that "it is {not} a problem we should expend too much effort on until and unless we start seeing actual cases where it causes problems." I think that we should remove the restriction; it is easy to deal with a few poorly named or spammish pages on patrol, and bad to frustrate innocent users with a misleading and likely irrelevant error message.--Prosfilaes (talk) 01:16, 10 December 2019 (UTC)

Request for an interface admin to edit MediaWiki:Gadget-ocr.jsEdit

Hi!

Since hOCR is currently buggy, please see mul:Wikisource:Scriptorium#Request_for_an_interface_admin_to_edit_MediaWiki:OCR.js : you can edit your local gadget to use fallback OCR as default one, just commenting the if condition in hocr_callback() function. —Pols12 (talk) 12:58, 21 December 2019 (UTC)

Sockpuppets of User:LupsteEdit

User:Lupste has been globally locked. Suspected sockpuppets of this account that have since appeared on Wikisource:

--EncycloPetey (talk) 20:15, 23 December 2019 (UTC)

Poems of Sappho and other works deleted as copyvio which have meanwhile slipped into PDEdit

The Poems of Sappho were deleted in 2013 as copyvio and moved to Wikilivres (which has unfortunately disappeared). Looking at Author:Edwin Marion Cox the work was published in 1924 and so it should be in PD now. Is it possible to restore it? It would be also great if it were possible to find other deleted works of this kind which later slipped into the PD and restore them too. --Jan Kameníček (talk) 14:23, 15 February 2020 (UTC)

As I commented elsewhere, our copy of The Poems of Sappho was a tiny fragment of the whole work, and not the first part either. Wikisource standards have changed, and we wouldn't host such an extract these days. For that work, it would be advisable to start from scratch. --EncycloPetey (talk) 17:21, 15 February 2020 (UTC)