[00:43:37] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [02:13:41] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [02:22:08] (03CR) 10Krinkle: Use null coalescing operator (031 comment) [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461737 (owner: 10Awight) [03:43:45] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [05:13:49] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [06:00:10] o/ [06:20:55] 10Scoring-platform-team (Current), 10ORES, 10User-Ladsgroup: Implement support for whitelisting and proxy requests for poolcounter in ORES - https://phabricator.wikimedia.org/T201826 (10Ladsgroup) a:03Ladsgroup It's done :] [06:43:54] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [07:25:18] 10Scoring-platform-team (Current), 10ORES, 10User-Ladsgroup: [Spec] Tracking and blocking specific IP/user-agent combinations - https://phabricator.wikimedia.org/T137962 (10Ladsgroup) a:03Ladsgroup With {T160692} implemented, this is basically done. [08:13:57] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [08:38:45] 10Scoring-platform-team (Current), 10Bad-Words-Detection-System, 10revscoring, 10Patch-For-Review, and 2 others: Add language support for galician - https://phabricator.wikimedia.org/T201142 (10Ladsgroup) https://github.com/wikimedia/revscoring/pull/411 [08:41:33] wikimedia/revscoring#1527 (galician - 2b6bcad : Amir Sarabadani): The build passed. https://travis-ci.org/wikimedia/revscoring/builds/432371550 [09:10:47] 10Scoring-platform-team, 10ORES: ORES uwsgi logs in logstash are useless - https://phabricator.wikimedia.org/T205256 (10Ladsgroup) [09:44:01] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [09:55:20] I won't be around for the next four hours, will be back after that [09:55:31] Relocating to Poland [11:14:05] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [12:44:09] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [13:42:02] Having trouble with my laptop. Will connect shortly [13:46:28] o/ [13:46:45] Still struggling with my laptop. Lenovo's docks seem to be... sub-par. [14:14:12] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [14:28:25] Good morning [14:32:53] o. [14:32:57] *o/ [14:33:05] * halfak sends an email to oit [14:38:03] halfak: apparently Q2 goals are due today; last week we worked on a list of goals and should be discussing them at today's meeting: https://etherpad.wikimedia.org/p/scoring_201819_q2_goals [14:38:15] Yikes. Thanks. [14:38:20] Glad y'all are on that. [14:39:40] I might need to push our sync meeting back because of someone trying to schedule something with me over the top of it. [14:39:59] I don't understand why anyone would think that our one weekly sync meeting is a good time to schedule over. [14:43:43] Anyway, I'll be there for this discussion. If we can't push our sync meeting back then I'll have to force them to reschedule the other thing. [14:50:46] I just added some notes so I know what y'all are thinking. [14:53:02] I also created https://www.mediawiki.org/wiki/AI_Community [14:53:12] With a goal of creating a central point for general discussions about AI in the Wikimedia communities [15:04:05] Related: https://meta.wikimedia.org/wiki/Advanced_technology#People_who_are_interested [15:04:25] I originally created this AI Community page on Meta, but I know your opinion of Meta. [15:04:49] well, I think that if we're trying to work with newcomers, meta admins seem to hostile towards that. [15:04:59] I'm not sure if we're trying to do something with newcomers here. [15:05:32] If we create the system on Meta, we get to be part of a broader portal called "Connect" [15:05:40] https://meta.wikimedia.org/wiki/Connect [15:06:24] And we get to make use of nice JS gadgets [15:06:46] https://tools.wmflabs.org/pageviews/?project=meta.wikimedia.org&platform=all-access&agent=user&range=latest-20&pages=Connect [15:06:48] :P [15:07:11] hey, more than i expected :D [15:09:27] o/ awight [15:09:45] halfak: o/ welcome back! [15:09:55] Reading backscroll--I'm fine with any time for our sync meeting [15:10:22] kk [15:10:30] Amir1: around today? [15:10:30] Is Amir1 out today? [15:10:33] halfak: I also finished up mapping use cases to software features in https://www.mediawiki.org/wiki/JADE/Use_cases and I am interested in writing a more detailed document that maps out different scenarios [15:10:35] ha [15:11:01] I'm around [15:11:06] Relocating [15:11:19] :D Amir1 Feel like doing some code review... [15:11:39] It's getting quite constipated, https://gerrit.wikimedia.org/r/#/projects/mediawiki/extensions/JADE,dashboards/default [15:12:19] Before that, let's talk about sync meeting timing. [15:12:56] Amir1, I had something scheduled over in a silly way. I'd like to push our sync meeting back. Maybe to tomorrow... :\ [15:13:25] Arg this is ridiculous. looking at my calendar, I had our sync meeting scheduled over as well as the hours afterward by the same person >:( [15:13:36] * awight grimaces at halfak's calendar [15:13:37] Or maybe we could just sync right now. [15:13:43] I'm around [15:13:47] I'm here [15:13:55] Amir1, would that work for you? [15:14:05] halfak: Your poor calendar has 30 minute breaks, that looks awful [15:14:34] Yes. :( [15:14:57] Not now for me [15:15:06] I'm on phone [15:16:09] I can be on pc in half an hour [15:16:15] If that helps [15:16:52] Yes. That is helpful. [15:17:23] Let's talk about the other Q2 things ASAP and then Amir1 can join to talk about logging. [15:17:32] I'm in the meeting call now. [15:17:32] Amir1: You could call in if not driving :) [15:17:33] https://hangouts.google.com/hangouts/_/calendar/YWFyb24uaGFsZmFrZXJAZ21haWwuY29t.75jfkuclhg9bjjoo0rsjo6an40?authuser=0 [15:18:11] The data here is as expensive af, 10 euro per one gig [15:18:43] gross [15:18:54] I meant voice call, but Google probably doesn't make that fun either [15:19:07] halfak: trying to join, weird glitches [15:19:10] kk [15:44:16] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [15:45:58] 10Scoring-platform-team, 10ORES, 10Patch-For-Review: Consistent TimeoutErrors when using Celery 4 - https://phabricator.wikimedia.org/T179524 (10awight) There's a Celery bug which seems to map to our issue, but I can't find it in the task history. When we do an experimental upgrade, we must have monitoring... [16:01:44] awight: will we have the sync meeting? [16:02:10] Amir1: no. Feel free to use the time for CR :D [16:02:19] :))) [16:02:20] sure [16:02:49] Cool, thank you! [16:03:01] I'm about to wrap up the final patch BTW [16:03:10] Just need to write maintenance scripts. [16:03:14] No peeking ;-) [16:06:21] awight: Going from the bottom. https://gerrit.wikimedia.org/r/q/project:mediawiki%252Fextensions%252FJADE+status:open [16:06:32] This still needs a small fix: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/JADE/+/448085 [16:06:36] Let me just do it [16:08:03] no, it has merge conflict now [16:08:28] (03CR) 10Ladsgroup: [C: 04-1] "The issue is not resolved and it has merge conflict, please fix." [extensions/JADE] - 10https://gerrit.wikimedia.org/r/448085 (owner: 10Awight) [16:09:01] (03CR) 10Ladsgroup: [C: 032] Drop "append-only" support [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461824 (owner: 10Awight) [16:10:08] (03CR) 10Ladsgroup: [C: 032] MCR API compatibility [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461826 (owner: 10Awight) [16:10:50] (03CR) 10Ladsgroup: [C: 032] Drop page judgments for this release [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461825 (https://phabricator.wikimedia.org/T202596) (owner: 10Awight) [16:12:07] I'll wait to rebase anything until the smoke clears... a lot happened. [16:12:32] (03CR) 10Ladsgroup: [C: 032] Make sure we assert things in tests [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461828 (owner: 10Awight) [16:15:52] (03CR) 10Ladsgroup: [C: 032] Update judgment content schema [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461255 (owner: 10Awight) [16:16:52] (03CR) 10Ladsgroup: [C: 04-1] Use null coalescing operator (031 comment) [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461737 (owner: 10Awight) [16:17:11] i think that would help for now [16:17:21] I will be back soon-ish [16:17:27] can't say for sure tbh [16:17:40] today is a little bit chaotic [16:18:03] Thanks for chipping away at it! [16:18:35] 10Scoring-platform-team (Current), 10ORES, 10User-Ladsgroup: ORES requests for wikidatawiki models=damaging end up with HTTP request timed out - https://phabricator.wikimedia.org/T204961 (10greg) p:05Triage>03High >>! In T204961#4602109, @Ladsgroup wrote: > I don't think it should be a blocker to the tra... [16:20:48] (03CR) 10Awight: Use null coalescing operator (031 comment) [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461737 (owner: 10Awight) [16:20:52] (03Abandoned) 10Awight: Use null coalescing operator [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461737 (owner: 10Awight) [16:25:55] awight: do you have the etherpad for scoring platform/research sync? [16:26:30] hared: https://etherpad.wikimedia.org/p/scoring_platform [16:26:33] harej ^ [16:26:40] for the sync with research? [16:26:42] ooh [16:26:56] https://etherpad.wikimedia.org/p/scoring_research_hangout [16:27:01] Thank you [16:32:13] (03Merged) 10jenkins-bot: Drop "append-only" support [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461824 (owner: 10Awight) [16:32:15] (03Merged) 10jenkins-bot: Drop page judgments for this release [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461825 (https://phabricator.wikimedia.org/T202596) (owner: 10Awight) [16:32:19] (03Merged) 10jenkins-bot: MCR API compatibility [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461826 (owner: 10Awight) [16:32:21] (03Merged) 10jenkins-bot: Make sure we assert things in tests [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461828 (owner: 10Awight) [16:32:23] (03Merged) 10jenkins-bot: Update judgment content schema [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461255 (owner: 10Awight) [16:38:29] (03CR) 10jenkins-bot: Drop page judgments for this release [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461825 (https://phabricator.wikimedia.org/T202596) (owner: 10Awight) [16:40:57] Amir1: are you around? [16:41:00] (03CR) 10jenkins-bot: Drop "append-only" support [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461824 (owner: 10Awight) [16:42:45] (03CR) 10Awight: [WIP] Secondary indexes for JADE pages (032 comments) [extensions/JADE] - 10https://gerrit.wikimedia.org/r/456078 (https://phabricator.wikimedia.org/T203037) (owner: 10Awight) [16:43:16] stephanebisson: I think Amir1 is commuting for a bit longer. [16:43:26] (03CR) 10jenkins-bot: Make sure we assert things in tests [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461828 (owner: 10Awight) [16:43:32] awight: ah ok, thanks [16:43:37] np! [16:44:24] awight: Do you know what needs to be done for a new model (articlequality) to be inserted into ores_model? [16:44:59] I thought it was automatic because the model is defined in config but apparently not. [16:45:11] stephanebisson: I do, more or less. Which wiki is this? [16:45:31] betalabs enwiki [16:45:45] looking... [16:45:57] (03CR) 10jenkins-bot: MCR API compatibility [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461826 (owner: 10Awight) [16:46:26] $wgOresModels looks correct [16:47:22] The model is provided by the server, https://ores-beta.wmflabs.org/v3/scores/enwiki/ [16:47:53] stephanebisson: I think the problem must be in the component responsible for surfacing the model in the UI. Is that the part you're working on now? [16:48:21] --cos I'm not aware of any UI for that yet, it won't show up in new RCFilters for instance. [16:48:53] awight: I've renamed wp10 to articlequality in PageTriage and it errors out as follows: [16:49:06] Exception caught: No model available for [articlequality] [16:49:19] happening in `ORES/includes/Storage/SqlModelLookup.php(42)` [16:49:23] Okay I think I have a clue. [16:50:01] We have some code which updates the ores_models table in the background, and it's possible that articlequality won't hit that code path. [16:50:26] Try making an edit to any page, actually... [16:50:37] oh, good call [16:50:56] On my way, will get on a pc in twenty minutes [16:51:00] (03CR) 10jenkins-bot: Update judgment content schema [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461255 (owner: 10Awight) [16:51:09] stephanebisson: yes, try editing [16:53:24] fyi https://github.com/wikimedia/mediawiki-extensions-ORES/blob/master/includes/Hooks/RecentChangeSaveHookHandler.php [16:53:52] 10Scoring-platform-team, 10ORES, 10Quibble, 10Browser-Tests, and 5 others: Quibble must include tests/selenium/LocalSettings.php (was ORES webdriver.io selenium test fail on CI due to lack of ORES server) - https://phabricator.wikimedia.org/T198201 (10hashar) a:05hashar>03None The test has been backed... [16:54:20] * awight scratches head [16:55:27] ah here's what I was looking for, [16:55:33] stephanebisson: FYI https://github.com/wikimedia/mediawiki-extensions-ORES/blob/master/includes/ScoreFetcher.php#L107 [16:55:59] awight: are you going to be joining in the scoring sync with research in five minutes? [16:56:10] Whenever a score is fetched, we check the ores_model table and update or upsert if necessary. [16:56:16] harej: yep [16:56:29] Thanks for the notification! [16:57:05] Right... I made a change, same error. I'll try again on the main namespace [16:57:14] harej: Will you? I see you have some agenda items [16:57:58] stephanebisson: +1 the articlequality model is only enabled on two namespaces it seems, sorry I didn't mention earlier: [16:58:01] 'articlequality' => [ 'enabled' => true, 'namespaces' => [ 0, 118 ], 'cleanParent' => true ], [16:58:02] I will be! [16:58:21] stephanebisson: you can query ores_model directly if that helps debug [17:14:19] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [17:35:00] Amir1: We just used one of the patches you merged in a conversation, thanks again! [17:38:51] stephanebisson: Any luck after making the main namespace edit? [17:39:32] awight: I've made several edits and still have the same problem... [17:39:55] I was editing draft before, which is 118, and should have worked [17:40:47] okay thanks, I'll create a bug [18:06:16] 10Scoring-platform-team (Current): Beta not updating ores_model for the article quality model. - https://phabricator.wikimedia.org/T205317 (10awight) [18:06:30] 10Scoring-platform-team (Current), 10Growth-Team, 10MediaWiki-extensions-PageCuration: Beta not updating ores_model for the article quality model. - https://phabricator.wikimedia.org/T205317 (10awight) [18:08:31] Amir1: ^ I'll see what progress I can make [18:09:03] Were a number of ORES models (particularly goodfaith models) rebuilt recently? fiwiki goodfaith for example is really bad now and didn't use to be [18:09:32] https://ores.wikimedia.org/v3/scores/fiwiki/?models=goodfaith&model_info=statistics.thresholds.false.%22maximum%20recall%20@%20precision%20%3E=%200.15%22 returns no results, that didn't use to be the case. Looking through https://ores.wikimedia.org/v3/scores/fiwiki/?models=goodfaith&model_info=statistics.thresholds.false the precision/recall numbers are all really low [18:11:19] I originally discovered this because https://fi.wikipedia.org/wiki/Toiminnot:ORESModels was reported to throw errors. The Recentchanges config was not updated for these model changes, so now it's defining threshold criteria that are impossible to meet [18:12:09] A few other wikis have these issues too. So far I've found (through error logs): fiwiki goodfaith, hewiki goodfaith, fawiki damaging, ruwiki goodfaith, srwiki goodfaith [18:18:18] ( cc awight ---^^ ) [18:21:27] 10Scoring-platform-team (Current), 10Growth-Team, 10MediaWiki-extensions-PageCuration: Beta not updating ores_model for the article quality model. - https://phabricator.wikimedia.org/T205317 (10SBisson) I've run `mwscript extensions/ORES/maintenance/PopulateDatabase.php --wiki enwiki -n 1 -b 1` to try to for... [18:22:49] RoanKattouw: Thanks for the alert! We should have an automated system for this kind of eyeballing, at model training time even. [18:23:12] The reason we don't catch it is that we're often bulk updating models, as you can imagine. [18:23:14] Yeah. As it happened we had a semi-automated one because of these notices in my code [18:23:22] oh neat! [18:23:53] Well, don't celebrate too soon, Timo is making me fix the notices because of the resulting log spam [18:24:05] hahaha fair enough [18:24:12] We should reevaluate the configs for all models that were updated (or really just all models) [18:24:33] I might try to do that today if this airplane wifi is sufficiently cooperative [18:24:54] stephanebisson: there is an ores_model row for articlequality already, I see [18:25:14] RoanKattouw: expenses-qualified :) [18:25:30] Not that I would have expected you to be doing anything other than working on a flight ;-) [18:25:32] It's JetBlue so the wifi is free :) [18:25:41] But also pretty slow and lossy [18:26:31] awight: Yes, there is one now [18:26:32] I appreciate that they're focused on keeping you in defiance of such a decisive and yet still poorly understood force as gravity. [18:26:55] stephanebisson: your use case is still crashing though? [18:27:16] Maybe jobs didn't run?... [18:27:20] awight: The app is working now. [18:27:29] okay, I'll note on the task [18:27:37] I wonder if it was added as a result of me running that maintenance script [18:27:59] 10Scoring-platform-team (Current), 10Growth-Team, 10MediaWiki-extensions-PageCuration: Beta not updating ores_model for the article quality model. - https://phabricator.wikimedia.org/T205317 (10awight) The app is working now, so maybe the failure was in initially running the job? [18:28:19] This should be happening transparently on a wiki with any non-zero activity, so appreciate the bug report either way. [18:28:26] Sorry for the block-shapedness! [18:29:20] (03PS1) 10Catrope: SpecialORESModels: Fix notices when ORES thresholds are broken [extensions/ORES] - 10https://gerrit.wikimedia.org/r/462554 (https://phabricator.wikimedia.org/T205228) [18:29:33] We expect the PageTriage code that uses articlequality to ride the train this week. So if we deploy config to enable this model today, we're confident that it will exist in ores_model soon enough? [18:29:58] For sure, although let's deploy code-first and then configure? [18:30:06] Or... just check production for ores_model entries. [18:31:58] "wp10" and no articlequality, for now. [18:32:24] I'm thinking the easiest migration is to enable both wp10 and articlequality for a cutover period. [18:32:38] That's what we did on beta and it didn't work too well. So I'm thinking we can deploy `articlequality => [ enabled => true ]` for it to initialize itself and when that looks good, we deploy the code that uses it [18:32:48] ah +1 in this case, ty [18:33:01] I'm interested in what went wrong, if you find notes some day... [18:33:12] the server should support both [18:33:30] (confirmed) [18:33:44] Amir1 was not too sure about enabling both since it duplicates data but doing it for about 1 day should be fine... [18:34:06] RoanKattouw: I love this BTW, thanks again for the patches! https://fi.wikipedia.org/wiki/Toiminnot:ORESModels [18:34:11] 10Scoring-platform-team, 10ORES: Updated ORES models can no longer satisfy configured threshold requirements - https://phabricator.wikimedia.org/T205320 (10Catrope) [18:34:30] awight: For sure! Also I just (---^^) filed a task for what we talked about [18:34:31] * awight ?uselangs for satisfaction https://fi.wikipedia.org/wiki/Toiminnot:ORESModels?uselang=fi [18:35:24] awight: This is the patch I would like to deploy https://gerrit.wikimedia.org/r/#/c/operations/mediawiki-config/+/461715/ [18:36:00] It mostly disable wp10 and enable articlequality except for euwiki because I don't know what using wp10 there [18:37:34] stephanebisson: Go ahead with euwiki IMHO, unless you think it'll reflect badly on page triage or something? [18:38:51] PageTriage is not on euwiki. I mentioned it because it's the only wiki with wp10 enabled and I have no idea what it is used for over there. [18:39:44] Ok then, if you think there's no consumer on euwiki I'll migrate it at the same time. [18:41:39] (03CR) 10Krinkle: SpecialORESModels: Fix notices when ORES thresholds are broken (031 comment) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/462554 (https://phabricator.wikimedia.org/T205228) (owner: 10Catrope) [18:42:18] I'll deploy it today during evening swat [18:44:24] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [18:45:05] stephanebisson: As far as I'm aware, you were the first consumer of wp10 [sic. :], if you know of others then it shows how out-of-date my understanding is (quite likely!). [18:48:26] 10Scoring-platform-team, 10ORES: Updated ORES models can no longer satisfy configured threshold requirements - https://phabricator.wikimedia.org/T205320 (10awight) Adding to @Catrope, we could add a step to the Makefile after writing model stats which would look for obvious shortcomings in the models built and... [18:50:40] halfak: I wanted to make a superficial change to the content schema, and can't remember if we went in circles about this yet: s/schema/scale/ ? [18:51:11] I understand that it implies a scalar so e.g. drafttopic is an awkward fit, bit IMO this communicates the meaning of the field very well. [18:51:43] *"scale" implies [18:52:45] I suppose even drafttopic will be a vector of scalars, since the topics will be from a static set. [18:57:36] (03PS2) 10Catrope: SpecialORESModels: Fix notices when ORES thresholds are broken [extensions/ORES] - 10https://gerrit.wikimedia.org/r/462554 (https://phabricator.wikimedia.org/T205228) [18:59:47] awight, Amir1 : I'm trying to run `ORES/maintenance/BackfillPageTriageQueue.php` in belatabs/enwiki as a trial run since we need to do it in prod soon. It runs a little then invariably errors out with a timeout. I'm wondering if this can be related to the throttling per-ip put in place recently, or if I'm asking too much of https://ores-beta.wmflabs.org/ [19:00:25] it is sending batches of 50 rev ids [19:01:12] stephanebisson: it can't be related [19:01:22] Ores returns 429 now [19:01:28] Not timeout [19:01:41] (Eating dinner atm) [19:01:50] ah ok, thanks [19:02:06] 10Scoring-platform-team, 10Growth-Team, 10MediaWiki-extensions-ORES, 10MediaWiki-extensions-PageCuration: ORES "is model really enabled?" PHP API - https://phabricator.wikimedia.org/T205323 (10awight) [19:02:58] ^ If we implement that, we can experiment again with ores_model by deleting the articlequality row on beta. [19:45:01] meetingmeetingmeeting [19:46:20] awight, I didn't understand that last specification you sent to Prateek. [19:46:32] "No schema can appear in any two judgments, regardless of whether the values agree or disagree, unless exactly one of the two judgments is preferred."? [19:47:21] halfak: Here's an implementation to make it more concrete, [19:48:05] https://gerrit.wikimedia.org/r/#/c/mediawiki/extensions/JADE/+/461736/ [19:48:37] I took this route because we were underspecified for cases like {damaging: true} vs {goodfaith: true}. [19:49:38] awight, it's really hard to read through your rules. But in this case, I'd say that those should be not allowed. [19:50:10] a judgment that doesn't include the full info is not allowed. [19:50:23] Not sure why we were considering allowing that at all in the first place. [19:50:45] Yes, diffs are hard to read, but the doc block summarizes it. [19:50:52] about this other point, though [19:51:15] show me an example of what you're talking about. "{goodfaith: true}" with no damaging judgment is not a thing? [19:51:31] Shouldn't it be impossible to have a judgment without it being marked "preferred"? [19:52:25] Regardless of whether there are one or more judgments? [19:52:31] That's one possible implementation, but it's an annoying restriction to have on the content, so I'm letting the first judgment in each schema be implicitly preferred. [19:52:50] BTW I realize now that we're still underspecified [19:53:07] awight, why not just enforce it. You're forcing future consumers to also encode that assumption. [19:54:07] [just to clarify, not to contradict] what do you think of the use cases as I sent them to prtksxna? [19:54:30] meh that will expand our scope here [19:55:23] The two alternatives we're discussing are A) client must include "preferred: true" when submitting an initial judgment [19:55:43] (03PS1) 10Sbisson: Backfill PageTriage: retry failed requests [extensions/ORES] - 10https://gerrit.wikimedia.org/r/462566 [19:55:44] vs B) client doesn't have to do anything with the preferred flag until submitting a second judgment. [19:56:23] I chose (B) but think that (A) is also reasonable because both put the burnden of setting preferred flag on the "contesting" editor. [19:56:46] (A) is nice because the editor is explicitly *changing* the flag [19:56:57] Right. So client must address preferred flags regardless. In one case, there's a meaningful assumption that requires counting while the other is just consistent. [19:57:12] At the time I implemented this, I imagined that both were annoying, but (A) more annoying to clients. Happy to review. [19:58:08] The submission of incomplete judgments seems problematic too. I wonder if we could look to the devs on our focus group to ask them about requiring a "goodfaith" assessment along with "damaging" [19:58:26] (03CR) 10Krinkle: [C: 04-1] "Sorry :)" (031 comment) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/462554 (https://phabricator.wikimedia.org/T205228) (owner: 10Catrope) [19:58:38] I don't imagine it would be a problem for huggler/stiki since the system is already designed to make the distinction [19:58:57] halfak: As for client logic, "counting" isn't as hard as you might be thinking. I'm imagining that the rules I chose allow a client to simply require a choice of "required" for the second, third, etc. judgments, so all the client has to know is that one judgment exists at all. [19:59:21] halfak: +1 let's get more opinions about damaging w/o goodfaith [19:59:28] Isn't it the way wikilabels works, for reference? [19:59:53] wikilabels will allow you to not submit one of the two fields but it will complain to you. [19:59:59] In practice, it never happens. [20:00:14] The tri-state feature seemed surprising to prtksxna at first, so maybe it's not a normal MediaWiki UX thing. [20:00:36] "never"? I've processed a lot of wikilabels data without one flag or the other. [20:00:48] I even wrote the processing that fixes it... [20:00:49] So here's what I imagine as a consumer for option (A): Get me the preferred judgment. For option (b): Get me the preferred judgment or if one doesn't exist, just get me the first (only) judgment. [20:01:12] With a comment next to it in the code saying, "If no judgment is marked preferred, then there should be only one and it is the preferred one." [20:01:29] awight, we used to not warn people and they could click save without selecting one. [20:01:31] :P [20:01:39] Some campaigns are very old. [20:01:44] ah got it [20:02:13] So it's something users are determined to do ;-) [20:04:18] It feels bad to require "preferred" on a single judgment because the meaning of a single judgment is unambiguous, and the rule I chose results in unambiguous data, simplified common-case data, and a tiny additional assumption to code into consumers. [20:04:45] that is a run-on but I stand by it :-p [20:05:12] * halfak goes to another meeting :( [20:09:13] bussing! [20:10:18] 10Scoring-platform-team (Current), 10ORES, 10User-Ladsgroup: ORES requests for wikidatawiki models=damaging end up with HTTP request timed out - https://phabricator.wikimedia.org/T204961 (10Ladsgroup) There are three ways to handle this, I'm planning to do all to make sure it won't happen again. Each of them... [20:14:28] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [20:57:43] heya halfak in https://ores.wikimedia.org/v3/scores/enwiki/?model_info=score_schema [20:57:53] should drafttopic schema say array of strings for prediction? [20:58:01] right now it says [20:58:02] "prediction": { [20:58:02] "description": "The most likely label predicted by the estimator", [20:58:02] "type": "string" [20:58:02] }, [20:59:22] eeek! good catch [20:59:42] post me a bug :D [20:59:46] ottomata, [21:00:06] k [21:01:46] https://phabricator.wikimedia.org/T205343 [21:01:47] :) [21:01:47] 10Scoring-platform-team, 10ORES: drafttopic score JSONSchema should specify array of strings for prediction type - https://phabricator.wikimedia.org/T205343 (10Ottomata) [21:05:51] 10Scoring-platform-team, 10ORES: drafttopic score JSONSchema should specify array of strings for prediction type - https://phabricator.wikimedia.org/T205343 (10Ottomata) Oh also, while you are at it "bool" is not a JSONSchema type. You mean "boolean". :) [21:13:49] halfak: I'm not sure you've stated your opinion of the AI Community page [21:14:27] should merge with the other AI community page. Otherwise it's great [21:15:32] Hmm. The other page is on Meta, and doesn't seem to be that active. [21:15:45] was recently created too [21:16:13] Also, do you have any good links I could use to fill in a News section? [21:16:35] Three or so, just to bootstrap [21:17:45] If we merge the two things together, should the combined page be on Meta or MediaWiki.org? [21:21:10] If Meta, we can just re-use the page I set up for m:Connect [21:22:36] https://meta.wikimedia.org/w/index.php?title=Connect%2FAI_Community&oldid=18323383 before I moved it to mw.org [21:25:20] * halfak is meeting for 30 more minutes [21:44:31] PROBLEM - ssh on ORES-redis02.experimental is CRITICAL: CRITICAL - Socket timeout after 10 seconds [21:53:18] is icinga2-wm a bot? [21:55:34] nevermind I found out it is [22:23:44] halfak: Backscroll includes a question about s/schema/scale/, wondering your opinion. [22:25:42] Amir1: Any more CR in ya? [22:25:59] I'll rebase over that rejected one-liner... [22:27:35] * awight cries for help [22:27:35] ! [remote rejected] HEAD -> refs/publish/master/secondary_schema (the number of pushed changes in a batch exceeds the max limit 10) [22:27:37] Yeah, so I don't see the argument for inconsistency. Sometimes "preferred" is required. Sometimes it isn't. When it isn't you're supposed to assume the right meaning. [22:27:57] I think it's more likely that clients will try to implement consistency (one judgment must be preferred) [22:28:28] (03PS6) 10Awight: Validate preferred flags [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461736 [22:28:30] (03PS6) 10Awight: Surface validation errors in the edit page [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461819 (https://phabricator.wikimedia.org/T197653) [22:28:32] (03PS4) 10Awight: Include annoying copyright preamble [extensions/JADE] - 10https://gerrit.wikimedia.org/r/460611 [22:28:34] (03PS4) 10Awight: Make articlequality scale configurable per wiki [extensions/JADE] - 10https://gerrit.wikimedia.org/r/461831 (https://phabricator.wikimedia.org/T203030) [22:28:36] (03PS3) 10Awight: Don't let subclasses couple to judgment storage [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462055 [22:28:38] (03PS3) 10Awight: Return Status objects rather than throw exceptions [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462034 [22:28:40] (03PS3) 10Awight: New classes for the judgment link table and to encapsulate judgment targets [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462053 [22:28:42] (03PS4) 10Awight: Wrap target entities in a JudgmentTarget [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462054 [22:28:44] (03PS4) 10Awight: Minor cleanup [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462056 [22:28:46] (03PS2) 10Awight: TitleHelper validates NS_JUDGMENT [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462151 [22:28:54] halfak: we can continue that, but I was asking about my other question, changing the "schema" field name to "scale". [22:29:01] (03PS2) 10Awight: Move ContentHandler package to Content [extensions/JADE] - 10https://gerrit.wikimedia.org/r/462152 [22:29:03] (03PS20) 10Awight: [WIP] Secondary indexes for JADE pages [extensions/JADE] - 10https://gerrit.wikimedia.org/r/456078 (https://phabricator.wikimedia.org/T203037) [22:29:28] I don't want to defend the preferred thing, it's just reasonable and already implemented. I'll just rewrite the code... [22:30:18] Oh I didn't see that one [22:30:20] 18:50 <+awight> halfak: I wanted to make a superficial change to the content schema, and can't remember if we went in circles [22:30:23] about this yet: s/schema/scale/ ? [22:30:26] 18:51 <+awight> I understand that it implies a scalar so e.g. drafttopic is an awkward fit, bit IMO this communicates the meaning [22:30:29] of the field very well. [22:30:34] 18:52 <+awight> I suppose even drafttopic will be a vector of scalars, since the topics will be from a static set. [22:30:43] * awight laments pasta formatting [22:31:14] [22:31:15] damaging/goodfaith are not scalar [22:31:20] O_o [22:31:24] I don't see the merits of "scale" [22:31:49] Also, I don't see "scale" in the schema. [22:32:05] My only point is that "schema" is extremely general and non-engineers will probably not immediately grok what it's about. [22:32:21] "dimension" [22:32:23] "scale" is a commonly used expression, e.g. Celsius scale, and would have the correct connotations IMO [22:32:28] that's overly technical again [22:32:35] though I like it :) [22:32:39] what's wrong with technical? [22:32:56] Ask 10 people what they think of when you say "dimension" :p [22:33:13] zero will say, a unit of analysis [22:33:24] https://www.google.com/search?q=dimensions+of+a+shirt [22:33:26] 9 will say "Star Gate" [22:33:35] 1 will say "t-shirt" lol [22:33:43] I think you watch too much scifi [22:33:44] :P [22:33:56] lol read my sci-fi thank you very much [22:34:08] T.V. sci-fi is usually just a war movie [22:34:27] https://www.google.com/search?q=dimensions+of+a+box [22:34:50] If we attempt to call something "dimension" we are immediately going to either win or lose the Internet on deployment day :p [22:34:59] "facet" "aspect" [22:35:08] What's wrong with "scale", again? [22:35:22] Booleans don't seem like a problem conceptually [22:35:34] I mean, it doesn't make much sense to me, I guess. [22:35:39] it's "cold" or "hot", never mind the centigrades [22:35:49] hmm kk, thanks for the feedback. [22:36:10] I can shelf that as something to test down the road if we care, though trying to clear these out to avoid data migrations. [22:36:36] As we've all said, Migration Will Happen (TM) [22:37:43] Amir1: 12 patches waiting for a happy resolution /o\ [22:38:59] I'm wondering if it would be rude to grab reviewers from other teams, to take the pressure off any bus factors of 1... [22:39:47] halfak: harej: Feel free to jump into https://gerrit.wikimedia.org/r/#/projects/mediawiki/extensions/JADE,dashboards/default at any point, even a conceptual/logical CR+1 would be helpful since that makes it easier to pull in reviewers from off the street. [22:42:51] There's lots of low-hanging fruit there. [22:44:45] harej: Did you have any thoughts about the "preferred" discussion in backscroll? Seems like you were suggesting the same logic as Aaron, right?