[06:12:04] [telegram] This is wonderful! I have some thoughts of how it could work, I can make a ticket around it. (re @fuzheado: Hi Susanna - I would be interested in helping, and maybe make this a good case study for getting people introduced to PAWS/Jupyter in a Wikimedia environment) [06:17:02] [telegram] I must say I don't know. The documentation is here https://wikitech.wikimedia.org/wiki/PAWS (re @psychoslave: Can the Wikimedia Jupyter instance run Julia?) [06:18:44] [telegram] I have tried to get started a couple of times. This could be the push I need. (re @fuzheado: Hi Susanna - I would be interested in helping, and maybe make this a good case study for getting people introduced to PAWS/Jupyter in a Wikimedia environment) [08:13:49] [telegram] If it’s really beginner friendly, aimed at non coders and is about reusing existing ones rather than starting new ones: same here (re @Jan_ainali: I have tried to get started a couple of times. This could be the push I need.) [08:15:44] [telegram] Do we have or could we create a collection of useful existing PAWS notebooks somewhere? Like we have example SPARQL queries? [08:54:36] [telegram] There is! https://wikitech.wikimedia.org/wiki/PAWS/PAWS_examples_and_recipes I don't know how seamless sharing is, but we can find out. (re @trnstlntk: Do we have or could we create a collection of useful existing PAWS notebooks somewhere? Like we have example SPARQL queries?) [08:55:26] [telegram] Yes we have! https://wikitech.wikimedia.org/wiki/PAWS/PAWS_examples_and_recipes I don't know how seamless sharing is, but we can find out. (re @trnstlntk: Do we have or could we create a collection of useful existing PAWS notebooks somewhere? Like we have example SPARQL queries?) [09:58:05] [telegram] Hi do you have a more detailed description of what you're looking to solve and I can perhaps start outlining a solution for PAWS/python? (re @Susannaanas: Yes we have! https://wikitech.wikimedia.org/wiki/PAWS/PAWS_examples_and_recipes I don't know how seamless sharing is, but we can find out.) [10:21:20] [telegram] I did this memo for spatial reconciliation. It's how I think it could be done, but there may be better ideas https://docs.google.com/document/d/1W8I3uUzjjXipNCY2JYj36867mCr1Fa9f7tfjmyKRA8U/edit# [10:59:03] [telegram] Ah thanks - do you have a full example case you can document in that Gdoc? (re @Susannaanas: I did this memo for spatial reconciliation. It's how I think it could be done, but there may be better ideas https://docs.google.com/document/d/1W8I3uUzjjXipNCY2JYj36867mCr1Fa9f7tfjmyKRA8U/edit#) [11:11:46] [telegram] I will work on it later today. I am working on a specific dataset and there's already some documentation in https://m.wikidata.org/wiki/Wikidata:WikiProject_Finland/Saami_place_names_to_Wikidata_as_lexicographical_data, may even too much. I will try to condense! [15:15:42] [telegram] Is there any chance we can create a custom Unix service on replica servers? [15:19:44] [telegram] Access /etc/systemd/system/ ? [15:21:22] [telegram] on which servers? [15:22:11] [telegram] replicas [15:22:59] [telegram] On the Toolforge DB replicas? No way (re @Yetkin: replicas) [15:23:30] [telegram] then, is there another way to create a system service there? (re @MaartenDammers: On the Toolforge DB replicas? No way) [15:23:31] [telegram] I think the answer will be “no” either way, but there are several kinds of servers that could be described as “replicas” [15:23:43] [telegram] why do you want to create a system service there? [15:24:01] [telegram] I have a long runnig job and I want to make sure it never stops [15:24:25] [telegram] Where are you running that now? Should be running on the Toolforge grid [15:24:36] [telegram] I usually start it with nohup but after some time (2-3 days) it stops [15:24:51] [telegram] What server is that running on exactly? (re @Yetkin: I usually start it with nohup but after some time (2-3 days) it stops) [15:24:56] [telegram] that is right (re @MaartenDammers: Where are you running that now? Should be running on the Toolforge grid) [15:25:28] [telegram] No, but we could easily add it. If there is demand. (re @psychoslave: Can the Wikimedia Jupyter instance run Julia?) [15:26:35] [telegram] that sounds like you’re running the job directly on the toolforge bastion? in which case, see rule 4 at [[wikitech:Help:Toolforge/Rules]] (re @Yetkin: I usually start it with nohup but after some time (2-3 days) it stops) [15:26:44] [telegram] on one of the DB replicas, I believe (re @MaartenDammers: What server is that running on exactly?) [15:27:08] [telegram] Can you do an uname -a ? [15:27:52] [telegram] uname -a gives [15:27:52] [telegram] Linux tools-sgebastion-07 4.19.0-0.bpo.14-amd64 #1 SMP Debian 4.19.171-2~deb9u1 (2021-02-08) x86_64 GNU/Linux [15:28:05] [telegram] yup, you’re on a bastion host [15:28:21] [telegram] You should be using jsub to submit your jobs [15:28:25] [telegram] all my cron jobs are there [15:28:59] [telegram] If I do that, can I make sure my job never stops? (re @MaartenDammers: You should be using jsub to submit your jobs) [15:29:39] [telegram] You can do something like this. That's one of my upload bots that runs most of the time: [15:29:40] [telegram] 0 0,4,8,12,16,20 * * * jsub -once -mem 3000m -j y -o /data/project/geograph/logs/geograph_uploader_resumed.log -N geograph_uploader_resumed /usr/bin/python3 /data/project/geograph/pywikibot/pwb.py /data/project/geograph/code/bot/commons/geograph_uploader.py -resumeupload >> /data/project/geograph/logs/crontab.log [15:31:12] [telegram] starting the job is not a problem. Just cannot make sure it will not stop [15:31:23] [telegram] any difference between starting it wi [15:31:36] [telegram] any difference between starting it with nohup and jsub? [15:31:42] [telegram] yes [15:32:36] [telegram] to clarify: to start the job, you’re just running nohup in the shell? [15:32:44] [telegram] to clarify: to start the job, you’re just running nohup ... in the shell? [15:33:02] [telegram] nohup directly on the toolforge bastion will run on a shared server meant to login, test and submit jobs, and is against the rules. Using it will degrade the experience for all toolforge users. [15:33:03] [telegram] I run this in the shell [15:33:04] [telegram] nohup php stt_surum.php & [15:33:43] [telegram] hmm, good to learn this now (re @chicocvenancio: nohup directly on the toolforge bastion will run on a shared server meant to login, test and submit jobs, and is against the rules. Using it will degrade the experience for all toolforge users.) [15:33:56] [telegram] back when it was my job I used to track users who do this and send messages telling them to use either jsub or k8s [15:34:52] [telegram] ok, I understand. Thanks for the response [15:35:12] [telegram] Manual is at https://wikitech.wikimedia.org/wiki/Help:Toolforge/Grid [17:52:27] [telegram] Re: accessibility, inclusivity and wikiBlind [17:52:28] [telegram] Hi all! Excited to be part of the group. Is anyone here interested in working on user creation problems with visual-only recaptcha? It stops anyone who can’t see from creating Wikipedia accounts. Urgently need help to update process [18:07:08] [telegram] recaptcha is Google's implementation of Captcha, we don't use that :) [18:07:16] [telegram] recaptcha is Google's implementation of a Captcha solution, we don't use that :) [18:08:15] [telegram] But yeah, the captcha we do use sucks for that reason and many others (none of this is new) [18:10:12] [telegram] @sam - definitely not new (legacy) and is still deeply problematic. [18:10:13] [telegram] hackers are needed! [18:10:15] [telegram] Ive got blind non-wikipedians who can help advise (from WikiBlind.org and afb.org) [18:10:49] [telegram] (thanks for reminding me of distinction between recaptcha and captcha) [18:13:59] [telegram] It's not necessarily an easy problem to solve, and while I don't want to completely dissuade you from spending time on it, it may not be the best use of your effort; especially depending on the solution it may not be easy to get deployed to Wikimedia sites. Obviously it could still be used by other sites too. [18:14:00] [telegram] There is work underway to see if we can use hCaptcha instead - https://phabricator.wikimedia.org/T250227 but it is still a way off (technically, and community buy in etc) from going live [18:14:20] [telegram] can wiki restful api work with page id as input or just title ? [18:16:41] [telegram] I think it's only titles [18:21:07] [telegram] If anyone here inside Wikipedia circles is even slightly interested in helping make progress on recaptcha - we need insiders to help (re @tehreedy: It's not necessarily an easy problem to solve, and while I don't want to completely dissuade you from spending time on it, it may not be the best use of your effort; especially depending on the solution it may not be easy to get deployed to Wikimedia sites. Obviously it could sti [18:21:07] [telegram] There is work underway to see if we can use hCaptcha instead - https://phabricator.wikimedia.org/T250227 but it is still a way off (technically, and community buy in etc) from going live) [18:21:21] [telegram] Let me know? Dr.Mel.ganus@gmail.com [18:22:05] [telegram] There's definitely people with various interest, @gtisza is one (at least Captchas more generally) [18:25:46] [telegram] @drmelganus is https://www.hcaptcha.com/accessibility an acceptable solution? [18:27:05] [telegram] seems hcaptcha requires emails to keep captchas accessible, not sure how I feel about that. (re @tehreedy: It's not necessarily an easy problem to solve, and while I don't want to completely dissuade you from spending time on it, it may not be the best use of your effort; especially depending on the solution it may not be easy to get deployed to Wikimedia sites. Obviously it could still be used by other sites too. [18:27:06] [telegram] There is work underway to see if we can use hCaptcha instead - https://phabricator.wikimedia.org/T250227 but it is still a way off (technically, and community buy in etc) from going live) [18:27:37] [telegram] For the accessibility things? Yeah... [18:27:47] [telegram] There's potentially other ways around it, but the standards aren't standards yet [18:28:16] [telegram] (on the one hand, not sure how to make a secure accessible captcha without it. On the other, it is higher burden than we are placing on sighted users) [18:29:51] [telegram] (on the one hand, not sure how to make a secure accessible captcha without it. On the other, it is a higher burden than we are placing on sighted users) [18:31:34] [telegram] It's possible something like https://privacypass.github.io/ might help eventually [18:31:53] [telegram] But you're right. Putting in a TTS type audio captcha isn't going to help, as it's also pretty easily broken [18:33:34] [telegram] (and I'm not saying that's the solution that was necessarily being proposed) [18:34:01] [telegram] right now blind users must either use a captcha solver of some sort or ask another user to create the account and make edit/ confirm their user. So a self service captcha that requires email seems like a vast improvement in my view. And accessibility trumps open source in my book. [18:35:14] [telegram] If we can work toward having a captcha that could work for our deaf-blind peeps, that’s the ultimate goal. In the meantime anything (even just needs analysis and research) would be great. [18:36:02] [telegram] In WikiBlind we have Graham Pearce and Skylar Covich with admin príveles who can do account creation when we know someone who is stuck. [18:36:29] [telegram] We have the request account thing on enwiki for example (re @chicocvenancio: right now blind users must either use a captcha solver of some sort or ask another user to create the account and make edit/ confirm their user. So a self service captcha that requires email seems like a vast improvement in my view. And accessibility trumps open source in my book.) [18:36:43] [telegram] I presume you’ve seen the existing tasks? (re @drmelganus: If we can work toward having a captcha that could work for our deaf-blind peeps, that’s the ultimate goal. In the meantime anything (even just needs analysis and research) would be great.) [18:39:07] [telegram] I have not seen those tasks yet. Context: I’m an organizer for WikiBlind.org but not a programmer. Outward focus - mostly unsure about what’s been going on so far and where to look at who has been doing what. You could post links here or share with me directly here on telegram. Thank you!! [18:51:04] [telegram] as it happens, I was planning to work on that! (re @tehreedy: There's definitely people with various interest, @gtisza is one (at least Captchas more generally)) [18:51:38] [telegram] more details in a few days, but https://phabricator.wikimedia.org/T281397 has the basic idea [19:12:07] [telegram] @gtisza yay! Looking forward to collaborating!