[00:25:57] anyone around? [00:26:53] are wikimedia image URLs 'stable', i.e can you guarantee that no one will replace one picture of a specific car type with 'another better picture' (rather they'd upload the better picture as a new image, and just mark the car type as a category) [00:28:54] davr0s, Wikimedia Commons tries to consider external users when overwriting or moving files. Seehttps://commons.wikimedia.org/wiki/Commons:Overwriting_existing_files and https://commons.wikimedia.org/wiki/Commons:File_renaming for some more information [00:30:21] File description page URLs can be considered stable for files within the Commons scope https://commons.wikimedia.org/wiki/Commons:Project_scope [00:35:55] Image file URLs? or file description page URLs? [00:38:27] image file url [00:39:53] Hotlinked files don't tend to fail well, if the file is deleted, renamed, or vandalised, you've got nothing. [00:40:03] ok i see from your link that there *is* indeed the possibility of renaming (by stewards) [00:40:23] Only file mover is required to rename [00:41:00] would there be the ability to refer to an image by some unique ID (genereated from a checksum?) as well as the (changeable) filename [00:41:19] usecase: I am collaborating with someone else building a image labelling service, [00:41:33] i was interested in refering to wiki-media directly as an image store [00:41:48] the collaborator raised the concern that 'files might get moved..' [00:42:11] if the files could be refered to by a unique ID, that would be enough [00:43:46] Old versions have the ISO datecode prepended to the filename, but not current versions [00:44:08] But generally, the rules would suggest it shouldn't be a completely different photo [00:44:19] Usually, it's a touched up/cropped version of the original [00:44:28] thats enough to kill the application [00:44:48] the use case is drawing polygonal annotations, so cropping the image would destroy the alignment [00:45:02] so we'd have to copy the images out [00:47:01] is there any other demand for a non-destructive ID to image mapping (e.g. keeping the original before the crop, available through the ID) .... I hope I'm not the only person in the world interested in stable image store [00:47:27] I can see the need for filename improvement, would the system be able to keep the original filenames (like a wikipedia 'redirect') [00:48:00] Looks like prop=imageinfo can grab a sha1 checksum for the file [00:48:03] https://www.mediawiki.org/w/api.php?action=help&modules=query%2Bimageinfo [00:48:21] ok thats coool, [00:49:36] if I could do the revese: get an image by submitting a SHA-1 hash .. that would solve my problem [00:49:55] (i.e. i could then store the hashes as image identifiers) [00:50:09] i suppose you still might delete them though [00:50:50] (its this sort of thing https://www.mediawiki.org/w/api.php?action=help&modules=query%2Bimageinfo just to illustrate the interest further, if it's not already clear from what I've said) [00:51:03] ^ I think that specific project is tied to 'flikr' [00:51:56] they have 330K images, wikimedia is millions of images? imagine the value if those could be marked up [00:52:30] even when an image focusses on one subject, you can mark-up the components [00:52:46] more data is always better [00:54:26] how big is the entire mediawiki dataset? [00:54:53] Since you're talking files, it's complicated. [00:55:45] https://www.mediawiki.org/wiki/1.6_image_storage#Proposed_new that suggestion seems to addres the probem I'm talking about, is it WIP? [00:55:58] "Use a content hash (first 64 bits of MD5?) for identification" [00:56:39] That was proposed a decade ago, and never implemented. [00:56:47] oh :( [00:57:26] what was the blocker? or just lack of time to do it/low priority? [00:58:00] just that one part: find an image *by* it's MD5 - would solve my problem , i think [01:07:00] davr0s, Got it [01:07:24] https://commons.wikimedia.org/w/api.php?action=query&format=xml&titles=File:Commons%20File%20mover.svg&prop=imageinfo&iiprop=sha1 turns a description page name into a sha1 and https://commons.wikimedia.org/w/api.php?action=query&format=xml&list=allimages&aisha1=202f0cedd56ff8550af7fbcf835257bcf3e317f8 will convert back [01:07:43] ok thats encouraging [01:12:02] ok thats the *description page*, there's still the hazard of a re-cropped image, but we could store the *image* sha1 *aswell*, and just accept the hazard of losing some (and perhaps keep a local backup just incase but we dont need the highest res etc..) [01:12:13] thats certainly better than nothing [01:12:23] No, that's the sha1 of the image displayed on the description page. [01:12:29] oh cool [01:14:56] thanks for the information [01:15:43] Glad to help [10:38:31] 1. Let's say Username 'a' is titleblacklisted in wiki A. [10:38:31] 2. Username 'a' is created in wiki B. [10:38:31] 3. What happens when 'a' visits wiki A and SUL account creation is attempted? Account created? [10:45:02] well, tested it, account get created [10:46:42] afaik titleblacklist for usernames only works on meta now [10:46:56] to prevent sul fragmentation [10:48:34] heh [10:49:04] Let me test that too... [10:52:36] indeed, local TBL for username no longer works [14:56:01] Hello [14:56:04] Anyone awake? [15:05:56] Ask, if someone can help you they will [15:42:28] AntiComposite: I was trying to find out where to send a vocal sample for WikiSpeech purposes [15:42:30] ? [16:22:35] ShakespeareFan00: https://commons.wikimedia.org/wiki/Special:UploadWizard ? [16:23:13] more info: https://www.mediawiki.org/wiki/Wikispeech [16:45:30] andre__:yes.. but that page hasn't seen much update on progress [16:48:38] I also had some thoughts based on some rather specfic Wikisource use cases [16:48:44] Namely long-s [16:48:55] needs to be translated as 's' in speech [16:49:31] as well as varuiou dipthongs like ae,oe,fl,ll, and so on [16:50:20] ShakespeareFan00: yes.. and #wikimedia-tech is an even broader place to ask in on a Sunday instead of using the contact info methods for Wikispeech defined on that wiki page... [16:50:27] but up to you :) [16:52:19] andre__: Apologies [16:52:57] I also was trying to find someone that was technically able to set up a way of using e-speak to generate audio for web remotely [16:53:03] (Wikispeech is using Mary) [16:53:20] Espeak-NG doesn't run on my client end system [16:53:38] However if Wikimedia has set up a back-end to remotely serve audio files from TTS... [16:53:53] Is there any reason why it couldn't support other 'free' engines? [16:54:00] Espeak-NG is GPL [16:54:24] And has text based files for defining lanaguages/phoneme sets [16:54:27] see https://meta.wikimedia.org/wiki/Wikispeech for contact info [16:54:34] see https://www.mediawiki.org/wiki/Extension:Wikispeech for setting it up [16:54:51] you need to ask Wikispeech maintainers, not #wikimedia-tech , I am afraid. [16:55:13] it's currently not even deployed on Wikimedia sites (see IRC channel topic)... [16:55:32] If it's not deployed then apologies... [16:55:35] * ShakespeareFan00 out [21:55:05] What's wrong with the recent changes here: https://fi.wiktionary.org/wiki/Toiminnot:Tuoreet_muutokset?hidebots=1&hidecategorization=1&hideWikibase=1&limit=50&days=7&urlversion=2 I see no results [21:57:15] Third filter [21:58:47] no [21:59:22] if you change from 50 to 100 last results, and then back to 50, then it's not empty [22:00:27] Strange when did you first notice this Stryn [22:00:33] just now [22:00:44] I don't visit there often [22:03:37] (now off to sleep) maybe there's some deprecated code in local common.js?