[00:06:19] leila: salam khoobi? [00:06:24] Ironholds, just saw the tweet. Congrats dude! [00:06:35] I hope you get the best reviewers! [00:06:40] salaam Amir1. khoobam. How is it going? [00:06:41] That certainly won't be me. [00:06:59] Mostly because I have a COI ;) [00:07:43] not bad [00:07:53] working with halfak on ORES :) [00:07:56] it has been great [00:08:03] :D [00:08:15] we are working on making an extension for ORES [00:09:26] what kind of an extension Amir1? [00:10:26] something that marks bad edits [00:10:32] let people to filter out good edits [00:10:46] ah! got it. [00:10:48] mw:Extension:ORES [00:10:55] * leila checks [00:11:20] Amir1, btw, we should consider adding article quality predictions to the category pages. [00:11:26] Cool if it extends beyond the MVP. [00:12:09] hmmm [00:12:13] great idea [00:12:34] do you mean we tackle it later halfak? [00:12:56] halfak: btw, did you among your 4000 things have time to look at the 'productionizing research' thing dartar emailed about? [00:13:10] the current one is definitely wrong and I'm wondering if I should make trivial edits now or wait for you [00:13:15] since you def. have a better pic of it than me [00:13:49] Amir1, yeah. Later if it doesn't make sense now. [00:14:05] YuviPanda, is it on-wiki? [00:14:35] okay :) [00:14:42] halfak: it was on wikitech [00:15:09] linkey? [00:15:10] https://wikitech.wikimedia.org/wiki/Research#Productization [00:16:02] YuviPanda, not sure what you are thinking of changing, but I want to put cache and horizontal scalability in a stage 2. [00:16:09] Or maybe they should be in stage 0 [00:16:21] Logging should be stage 2 [00:16:35] I don't know why we have a stage 3, but there should be more than docs. [00:16:39] yeah [00:16:42] and docs shouldn't be stage 3 [00:16:55] Docs should probably be stage 1 [00:16:57] I def think cache and horizontal scalability have to be stage 0 [00:17:03] before you start coding even [00:17:12] since they're mostly dependent on the general shape of what yu build [00:17:13] Really awesome swagger docs could be stage 3 [00:17:14] than specifics [00:17:42] Why are you pinging everyone with an "a" name? [00:17:53] ? [00:18:02] YuviPanda, +1 for cache and scalability at stage 0 [00:18:22] halfak: I think the one hour we spent in the office many months back has paid itself back manyfold [00:18:37] and I think soon that one hour can be done by madhuvishy than me :D [00:18:45] YuviPanda, looking at the logs, my client is freaking out. BRB [00:18:58] haha [00:19:00] ok [00:20:31] YuviPanda: i was looking at that page - wikilabels can be added to that list too? [00:20:58] madhuvishy: +1 [00:21:06] adding [00:23:49] YuviPanda: I was wondering - where does the setup through docker/containers way play into this? [00:24:05] madhuvishy: I think it's only tangentially erlated [00:24:07] *related [00:24:07] do we wanna experiment with it more before making that an option? [00:24:14] hmmm [00:24:20] madhuvishy: yeah, a lot more [00:24:23] definitely [00:24:39] madhuvishy: I think it'll do the: staging env, health checks, deployment system thing there [00:24:41] that would make sense - it's taken a while for this setup to come this far [00:24:47] but conceptually we'll still need to do all the other things [00:24:51] yeah [00:24:51] it's just a matter of puppet vs something else [00:24:56] yup [00:24:57] and so for most people it doesn't change much at all [00:24:59] makes sense [00:28:52] madhuvishy: I think most important [00:28:55] is [00:28:57] > allow additional load be taken up by simply adding new instances [00:29:04] and > caching [00:29:04] Yup! [00:29:06] since [00:29:10] these are fundamental architectural things [00:29:22] and if you get those wrong it's a long way to come back from them [00:29:31] it may also be worth mentioning dependencies, and best practices around them? [00:29:33] true [00:29:39] dependencies in what sense? [00:29:44] debianizing/pip/venv etc [00:30:02] aah [00:30:04] hmm [00:30:06] so [00:30:09] that's the religious parts, almost [00:30:11] which is [00:30:13] what language to use :) [00:31:58] true - but in general, do we care about isolating them from the rest of the system [00:32:09] right [00:32:12] so that's an ongoing conversation [00:32:16] that is very much in flux [00:32:21] okay [00:32:24] so you're right that needs guidelines [00:32:30] except we don't have any guidelines to give people yet [00:32:51] may be we should have a WIP section? [00:33:01] yeah [00:33:04] +1 [00:39:30] halfak: YuviPanda I added a WIP section here - https://wikitech.wikimedia.org/wiki/Research - feel free to edit as needed [00:40:09] \o/ thanks [00:51:54] madhuvishy, looks good. I'd be bold with making edits there. [00:52:31] halfak: thanks, okay :) [00:52:31] YuviPanda, if you implement the adjustments you think are right, I'll support those edits [00:52:36] let me do it [00:56:47] halfak: madhuvishy I made some [00:56:57] stage 0 is all talk [01:06:56] * DarTar runs out to pick up the girls, ttyl everybody [01:08:14] wah he was here? [01:29:02] halfak: madhuvishy I think we should provide application structure templates [01:29:07] 'simple stateless webserver' [01:29:16] 'simple stateless webserver doing lots of network operations' [01:29:23] that would be cool [01:29:23] 'simple stateless webserver doing lots of CPU operations' [01:29:35] we can get to stateful at some point later [01:29:38] since we don't have any [01:29:41] atm [01:29:53] I don't know if we'll have a truly stateless one tho [01:31:17] madhuvishy: I think ores falls under 'stateless webserver doing lots of CPU operations' [01:31:27] (and hence the celery) [01:31:58] yeah [01:32:47] YuviPanda: when you say templates, what are you thinking? [01:32:59] madhuvishy: oh, I'm thinking of just a page defining things [01:33:01] not actual code [01:33:07] ah yeah okay [01:33:11] just [01:33:18] madhuvishy: basically [01:33:22] madhuvishy: step 0 [01:33:29] can be generalized for common use cases [01:33:36] yup [01:34:12] madhuvishy: I think ifttt is 'simple stateless webserver doing lots of network operations' [01:34:24] yeah! was just going to say [01:36:09] madhuvishy: I'm oscillating back and forth between doing it in puppet [01:36:12] madhuvishy: vs doing it in k8s [01:36:22] the latter is probably still not stable enough for iffttt [01:36:57] YuviPanda: it may be a good idea to set it up in puppet and then play with k8s? [01:37:06] that set up is not hard [01:37:16] madhuvishy: yeah [01:37:19] madhuvishy: I think so too now [01:37:24] madhuvishy: so we have a stable spot for it [01:37:27] yup [01:37:29] and then an unstable spot to play with it [01:37:42] +1 [01:38:06] madhuvishy: can you take up the puppetization? :D [01:38:12] YuviPanda: yeah [01:38:14] madhuvishy: I think maybe it doesn't even need a module. just a role might be enough [01:38:27] i was gonna do it anyway. i'll do it in the next couple days [01:38:33] madhuvishy: awesome [01:38:51] hmmm - yeah just uwsgi right? [01:39:10] i have to patch the actual code with a wsgi file [01:39:29] and change the fab file may be [01:39:36] madhuvishy: yeah [01:44:28] okay - i'll make a phab task. (i typed fab task first :/) [01:45:35] YuviPanda: do you know which project it'd go in? [01:45:50] halfak, thanks! [01:46:03] madhuvishy: let's make an iffttt one :D [01:46:29] YuviPanda: on phab? or labs? or both? [01:46:36] labs i know [01:46:43] madhuvishy: ah, I think ticket might be on github [01:46:46] and labs on phab [01:46:47] ah [01:46:50] yeah okay [01:49:58] YuviPanda: https://github.com/wikimedia/ifttt/issues/7 [01:50:39] \o/ [01:50:41] cool [01:51:01] madhuvishy: do you want to split that into: add wsgi file, and then fix fab? [01:51:16] Yup, those will be separate anyway? [01:51:24] puppet will be in gerrit [01:51:31] righy [01:51:32] *right [01:51:45] i'll make 2 more issues [01:51:47] one of the things I used to do once I was proficient in something [01:51:50] is to do a speed run :D [01:52:00] guess how much time it'll take you to do something [01:52:03] then set a timer [01:52:05] and do it [01:52:08] and see if you are far off [01:52:09] or close [01:52:16] and see why it went wrong [01:52:22] helps improve personal time estimation a lot [01:53:05] YuviPanda: yeah i think that'll help for this [01:54:33] +1 [01:58:03] YuviPanda: requested labs project and made 2 more issues on github [01:58:11] madhuvishy: \o/ thanks! [13:34:51] 10Quarry: Cannot download data from a query with Unicode characters in its title - https://phabricator.wikimedia.org/T123031#1947747 (10XXN) I confirm this. I encountered the same problem with http://quarry.wmflabs.org/query/6945 [14:44:14] o/ [15:23:51] \o [17:55:08] o/ schana [17:56:18] hi [17:56:40] Looks like we're on for Scrum of Scrums in a half hour. [17:56:55] hey schana! [17:57:27] I just noticed you don't have a calendar invite, so I added you. [17:57:38] thanks [17:58:00] I would have missed it otherwise [17:58:43] YuviPanda, can you tell me what the status of prod hardware for ORES is? [17:58:48] well [17:58:51] * halfak wants to add to SOS agenda [17:58:53] the oresredises are there [17:59:04] it just needs someone to do the actual installation and stuff now [17:59:15] blocked on ops time, definitely [17:59:18] rather than hardware now [17:59:33] Cool. Do you already have a card for this work or should I make one? [17:59:44] halfak: there's one [18:00:25] halfak: https://phabricator.wikimedia.org/T106867 [18:00:27] I think? [18:00:33] halfak: maybe needs more specific ones [18:08:28] YuviPanda, I think I can work with this card today. [18:08:46] I'd like to get a sense for what cards you would create to have more specific ones (since I imagine you have this setup plan in your head) [18:08:56] I'd be happy to write up those cards and link them together. [18:09:03] If you can give me a bulleted list. [18:13:16] halfak: sorry, in another meeting. [18:13:29] halfak: but I think the tasks linked under it are most of it maybe [18:13:33] the things left are: [18:14:37] * halfak prepares to start creating tasks [18:14:50] bah, actually have to pay attention in meeting [18:14:52] gimme 5min [18:15:00] no wortries :) [18:15:48] * halfak completes hyperparameter optimization for all eqitquality models for revscoring 1.0. Now for the article quality models. [18:21:11] schana, I've updated the SoS etherpad with some updates re. ORES. See https://etherpad.wikimedia.org/p/Scrum-of-Scrums [18:21:24] hi halfak. [18:21:50] Usually we make notes of what to report in the staff etherpad to make the SoS attending easier, but since I'm leading today, I just put them right in the SoS pad. (staff etherpad: https://etherpad.wikimedia.org/p/ResearchStaff) [18:21:59] Hey lzia! G'morning :) [18:22:14] I will be in the Reading team QR until 11:30. I will do my best to come out of the meeting 10 min earlier, to make it on time for A/V testing. If I can't make it, halfak, can you make sure we have a Jerome and his A/V as well as yours work? [18:22:36] Sure. I'll join up the full 15 minutes early. [18:22:44] perfect. thanks halfak. [18:22:45] Do you have contact info for Jerome? [18:23:09] And I also moved your abstract to the first time slot, so Jerome can see you in action once, before he gives his presentation [18:23:25] yes, halfak. he is in the Calender event we share [18:23:32] can you check and confirm, halfak? [18:26:49] lzia, confirmed [18:27:00] perfect. thanks, halfak. [18:32:15] halfak: https://phabricator.wikimedia.org/T106867#1948882 [18:32:40] <3 [18:34:03] np [18:34:07] * YuviPanda runs to office [19:22:25] ) | Channel is publicly logged @ http://bots.wmflabs.org/~wm-bot/logs/%23wikimedia-research [19:23:08] halfak: now glue that together [19:23:37] ) | Channel is publicly logged @ http://bots.wmflabs.org/~wm-bot/logs/%23wikimedia-research [19:23:50] Weird. It doesn't want to get the last bit. [19:24:09] aw, I can't set the topic [19:24:19] Ahh. I had a weird unicode space in there. [19:26:40] * Emufarmers clicks a John Oliver link on the sidebar while waiting [19:27:59] o/ Hey folks [19:28:04] hey all, we’re about to start with the research showcase https://t.co/tdESrvwEhd [19:28:20] I’ll be taking and relaying any questions from IRC [19:28:25] I'm presenting today, so send your questions to DarTar [19:28:30] :) [19:29:14] will DarTar run out of battery? [19:29:27] ^ looool [19:29:29] ;) [19:30:55] Nettrom: neverrr [19:31:15] Hi everyone. Welcome to the showcase. [19:31:22] Please ping me or Dario with questions! :-) [19:31:52] leila, do you have a link to the two decks handy? [19:33:11] no, DarTar. [19:33:23] alright, no worries [19:33:27] like, I don't have the link to the decks, not that I don't have them handy. [19:39:21] leila: where? [19:39:34] He said after the presentation, DarTar, right? [19:39:35] :D [19:39:38] ah cool [19:39:51] btw, DarTar, fyi, I didn't get your reminder email. [19:39:57] not to worry about it, just fyi. [19:40:29] oh, where? [19:40:31] leila: ^ [19:40:44] did you say you sent a reminder to wmfall, DarTar? [19:41:09] no I didn’t, I saw your email and I thought it was okay. Sent a reminder via social media [19:41:17] ah! cool! [19:44:28] wondering if a more apt comparison would be the rate of productive sessions [19:45:17] efficiency: also some possible factors related to cohorts [19:45:22] DarTar: another thing is the slope of the two plots, comparing 2006 and 2015. [19:45:45] The scales on the plots are different, so the green lines as slopes are not necessarily comparable. [19:45:59] fair [19:47:04] these are enwiki, DarTar? [19:47:09] yes [19:50:17] It's a good time to start collecting questions for Aaron. Please share them here if you have one or more. [19:51:04] I have 2 if nobody else has questions [19:52:21] or 3 actually :) [19:52:47] What does he use on his hair, and in his opinion as a professional scientist can people really not tell the difference between I Can't Believe It's Not Butter and actual butter? [19:53:10] Ironholds: I think you should walk up to the room and ask it. :D [19:53:27] hahah [19:55:17] thanks halfak: 1 minute warning [19:55:37] other questions from the audience, lmk before I ask mine [19:55:40] what was that self persisting token thingy? Is that survival of text written by someone ? [19:57:23] Is this content persistence thingy a tool I can run on myself and such? [19:58:21] Emufarmers: I’ll ask that Q, thx [19:58:27] (I was going to ask the same) [19:58:55] someone hugs him for me [19:58:58] that is awesome [19:59:16] I could not imagine one would be able to track such persistence of content over time / through other edits [19:59:18] I will ha ha [19:59:26] oh, I'm a little behind on the stream. I think he just indicated that was on the to-do list [19:59:47] * hashar hugs [20:00:59] ah some research group at IBM did something potentially similar https://www.research.ibm.com/visual/projects/history_flow/results.htm [20:01:12] visualization of sequence of strings over time [20:01:30] potentially leading to generate the actual list of authors for a given version [20:01:49] paper http://alumni.media.mit.edu/~fviegas/papers/history_flow.pdf [20:01:54] but you guys must be aware of it [20:02:11] ^ yes. One of the things I'd like to be able to do is have an uber, fast, Wikiblame. [20:02:36] yesssssssss [20:02:55] That would give authorship for articles and allow one to quickly find all content contributed by an author. One of the use-cases that I've learned about is dealing with hoax editors' contributions after they have been found out. [20:03:00] halfak: link to those slides, so I can share some highlights on @wikiresearch [20:03:07] ala git blame :-} [20:03:08] DarTar, uploading now. [20:03:12] Yeah. :) [20:03:16] thx [20:03:28] But it's hard. We'll be pushing the limits of what can be indexed. But I'm sure we can do it. [20:03:41] Q: who is speaking ? [20:04:09] halfak: Is your stuff generated from live data right now, or do you use dumps? [20:04:16] Dumps. [20:04:23] hashar: https://www.mediawiki.org/wiki/Wikimedia_Research/Showcase#January_2016 [20:04:42] Jerome Hergueux formerly at Sciences Po/Berkman [20:04:49] now at ETHZ [20:05:11] yet another french speaking doctor :D [20:05:36] hashar: I know, it’s getting uncomfortable :) [20:05:42] we’re surrounded [20:05:47] :D [20:06:06] time to change the official language [20:06:24] c'est bon, non? ;) [20:06:36] {soyons gras} [20:06:42] Nettreum: shhh [20:07:11] https://www.mediawiki.org/wiki/Wikimedia_Research/Showcase#January_2016 slides linked [20:07:21] https://commons.wikimedia.org/wiki/File:Anon_productivity_and_productive_efficiency_in_English_Wikipedia_%28Showcase,_Jan._2016%29.pdf [20:07:28] DarTar, ^ [20:07:57] halfak: thanks [20:08:38] My social type is awesome. [20:08:49] heh [20:09:06] so WMF research is in touch / partnership with universities around the world to conduct research aren't you? [20:09:25] hashar: this is a study from 2012 [20:09:28] but we like to be. [20:09:28] hashar, yes. As much as we can manage on top of our own work. :) [20:09:44] that is nice [20:09:56] will send you guys when asked :-D [20:09:57] yeah. if we can do more of it, hashar, that'd be great. [20:09:57] this game was played with real money, IIRC [20:10:18] hmm, what about "4. it's fun!" [20:11:29] Emufarmers: good point, survey research points to that as a motivator [20:12:01] Btw link to the paper: http://www.parisschoolofeconomics.eu/IMG/pdf/hergueux_paper-2.pdf [20:12:20] * guillom has been having it in his reading list for weeks, still hasn't read it. [20:12:58] * leila makes a note to read and discuss the paper with guillom. [20:13:50] schana: hey! [20:13:52] s [20:14:02] schana: there's a security issue I'm dealing with now, so will have to postpone our meeting :( [20:14:12] YuviPanda: okay [20:15:18] DarTar, Q: When looking for reverts with explanation, did you handle auto-generated edit summaries like those generated by rollback and undo? [20:15:39] schana: sorry about that! I'll move the meeting [20:15:44] halfak: logged [20:15:55] YuviPanda: no worries :) [20:16:37] hashar: https://www.mediawiki.org/wiki/Wikimedia_Research/Collaborators [20:16:46] * DarTar waves at schana [20:17:45] halfak: I’ll let you ask the Q directly since you’re in the hangout [20:17:53] DarTar, sounds good [20:18:26] did someone catch what the relationship was with age? [20:19:36] So, Barnstars... are they controlled for how many edits? [20:19:47] Maybe people who edit more get more barnstars and start more edit wars. [20:19:52] Because they edit more. [20:21:38] or maybe people starting edit wars do it to defend their point, eventually achieving a better quality articles which earn them the barnstar [20:21:40] * Emufarmers gives halfak a barnstar for being on my side in an edit war. [20:21:46] i.e. they fight for their articles [20:23:50] so A gives 10, B receives 30 and B can gives back 0-30 to A ? [20:23:52] Emufarmers: do you mean actual age or WP tenure? [20:24:36] yes, hashar. [20:24:37] actual age, I think. He said something about how it was the only demographic factor that showed a significant correlation [20:24:42] isn't that controlled for in the variable with disputed content proportion? (or whatever it measured) [20:25:01] just one note, hashar: that A can decide how much to give as well. So A gives a number between 0-10 [20:25:16] and why B has 10 as well ? [20:25:31] Admins who admin more see the worst of the world and lose trust? [20:25:53] Because if B sends more of his own money to A, A has more money that they can can send in the moneytriplator. [20:26:08] (I guess.) [20:26:15] hashar: it's probably to control for greed. [20:26:29] hashar: if you don't have anything, you are not starting at a level playing field. [20:26:32] Oh; good point. [20:26:36] :) [20:26:37] that is a loop so ? [20:26:49] money does A --- x3 --> B --> A --- x3 --> B [20:26:50] Maybe trust diversity is desirable. [20:27:20] halfak: or they don't change, and it's just that the trusting ones leave. [20:27:21] folks, ping me directly if you have a Q [20:27:21] merci Jerome ! [20:27:36] hashar: yes, [20:28:17] v good point halfak [20:30:49] great, looks like I missed one slide while multitasking :D [20:34:10] Maybe Jerome is talking about http://www.pensivepuffin.com/dwmcphd/syllabi/info447_au10/readings/kriplean.et.al.Articulations.CSCW08.pdf? [20:34:15] That's not Shaw/Hill. [20:35:02] \o/ [20:35:09] Thanks for coming! [20:35:57] Thanks everyone! [20:36:00] great presentations and discussion, thanks everyone! :) [20:37:08] thanks, Nettrom, and others. :-) [20:38:42] Thanks everyone for coming. :) [20:39:00] halfak: when you get a chance, please upload your deck. :-) [20:39:11] leila, already done and on the page :P [20:39:15] haha [20:39:15] :D [20:39:31] CC-BY-SAwesome [20:39:43] It's always a great feeling to get an opportunity to share research results with the community as opposed to traditional research seminars. [20:40:14] anytime SalimJah. ;-) [20:40:15] SalimJah, +1. We get to do a lot of imagining what research might be as Wiki people. [22:34:09] DarTar: are we meeting? [22:34:22] oops, bumped into toby and started trolling him [22:34:26] * DarTar coming [22:34:40] nothing is more fun than that. ;-) [22:37:45] heh [22:44:33] o/ DarTar [22:44:44] When you get a minute, could you review and leave any final comments on https://en.wikipedia.org/w/index.php?title=Wikipedia_talk:Labels/Edit_types/Taxonomy