[00:00:33] 10JADE, 10Scoring-platform-team (Current), 10Design: Come up with view mode for JADE pages - https://phabricator.wikimedia.org/T208819 (10awight) Oops! The text is not translatable yet, this is a bug. [00:02:43] 10JADE, 10Edit-Review-Improvements, 10Growth-Team: New watchlist filter: diffs without corresponding Judgment-space pages - https://phabricator.wikimedia.org/T209548 (10Harej) Pieces of data I am interested in: (1) What is the opt-out rate for new RC filters between new and experienced editors? I understand... [00:11:26] awight: we should really consider archiving our etherpads to officewiki; I don't know if etherpad has any availability guarantees [00:11:56] +1 also better for long-term collaboration, supports links and images... :) [00:12:40] Would you mind if I went ahead and did that? [00:14:07] Thanks x 100! [00:28:21] I tried copying and pasting a year's worth of notes to visual editor and it screwed up the indentation [00:28:30] And got rid of the strikethroughs [00:28:58] So as a medium step I'm moving the archives to separate etherpads, with cross-links [01:30:34] (03PS1) 10Awight: [WIP] Render a summary of the target wiki entity [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473652 [01:34:48] Unfortunate outcome, ^ I'll post a screenshot in a minute [01:35:11] 10JADE, 10Scoring-platform-team (Current), 10Design: Come up with view mode for JADE pages - https://phabricator.wikimedia.org/T208819 (10awight) {F27224625} [01:35:15] (03CR) 10jerkins-bot: [V: 04-1] [WIP] Render a summary of the target wiki entity [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473652 (owner: 10Awight) [01:35:39] (03CR) 10jerkins-bot: [V: 04-1] [WIP] Render a summary of the target wiki entity [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473652 (owner: 10Awight) [01:36:49] Now I think a diff line would be better than the nice RC output [01:37:10] *page history [01:39:35] 10JADE, 10Scoring-platform-team (Current), 10Design: Come up with view mode for JADE pages - https://phabricator.wikimedia.org/T208819 (10awight) A diff line might have better formatting, but still isn't quite right: > (cur | prev) 14:43, 22 September 2018‎ Admin (talk | contribs)‎ . . (243 bytes) (+243)‎ .... [01:42:31] * awight makes gurgling sounds at HTML concatenation in HistoryAction [01:52:23] Alas, I’m physically home but my mind is still in the office... [01:53:40] Is there any way we could make the diff show up? Rather than just the line in the RC feed [01:53:59] I think MediaWiki supports one column diffs now [01:54:36] But then, people are going to want the two column variant, and would probably not be happy to be told they only get one column [01:57:29] harej: Good thought, and the full diff output already includes well-formatted links. [01:57:50] 1- or 2- column is probably hidden by a service abstraction [02:00:01] (03PS2) 10Awight: [WIP] Render a summary of the target wiki entity [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473652 (https://phabricator.wikimedia.org/T208819) [02:01:57] harej: One thing I'd like to avoid, is the judgment controls appearing at the bottom of the page under a long diff. Any ideas how we can avoid that? [02:02:20] (03CR) 10jerkins-bot: [V: 04-1] [WIP] Render a summary of the target wiki entity [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473652 (https://phabricator.wikimedia.org/T208819) (owner: 10Awight) [02:02:43] err I think I'm physically working at home, but my mind is wandering. [02:03:16] This is fun though, and has a skunkworks edge. "what UI?" [02:03:58] (03CR) 10jerkins-bot: [V: 04-1] [WIP] Render a summary of the target wiki entity [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473652 (https://phabricator.wikimedia.org/T208819) (owner: 10Awight) [02:06:58] 10JADE, 10Scoring-platform-team (Current), 10Design, 10Patch-For-Review: Come up with view mode for JADE pages - https://phabricator.wikimedia.org/T208819 (10awight) @Harej is clarifying in IRC that "context" can be the full 1- or 2- column diff, not just a summary line. We need to design how this will fi... [03:25:59] o/ [06:25:34] 10JADE, 10Scoring-platform-team (Current), 10DBA, 10Operations, 10TechCom-RFC: Introduce a new namespace for collaborative judgments about wiki entities - https://phabricator.wikimedia.org/T200297 (10Marostegui) >>! In T200297#4745149, @awight wrote: > @Marostegui Pinging for review of these two files, h... [10:13:10] o/ [11:40:25] 10ORES, 10Scoring-platform-team, 10Patch-For-Review: Implement twemproxy for ORES production Redis - https://phabricator.wikimedia.org/T122676 (10Ladsgroup) >>! In T122676#4741880, @akosiaris wrote: > * Have a look at https://redis.io/topics/sentinel. Evaluate it, figure out what this means for the service a... [13:16:16] afk for lunch [13:16:19] will be back soon [14:02:51] back now [14:07:42] 10ORES, 10Scoring-platform-team, 10Patch-For-Review: Implement twemproxy for ORES production Redis - https://phabricator.wikimedia.org/T122676 (10Ladsgroup) I looked at sentinel. It's a little bit complex but easily doable. We probably need to do install redis on ores nodes and make them to run sentinel (Exa... [14:16:43] 10ORES, 10Scoring-platform-team (Current): Migrate ores celery configs to celery 4 - https://phabricator.wikimedia.org/T209587 (10Ladsgroup) [14:42:36] https://github.com/wikimedia/ores/pull/284 anyone to review? :( [15:21:41] (03PS1) 10Ladsgroup: Drop all of service integration with the API hook handlers [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) [15:24:26] (03CR) 10jerkins-bot: [V: 04-1] Drop all of service integration with the API hook handlers [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) (owner: 10Ladsgroup) [15:27:21] (03CR) 10jerkins-bot: [V: 04-1] Drop all of service integration with the API hook handlers [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) (owner: 10Ladsgroup) [15:36:13] (03PS2) 10Ladsgroup: Drop all of service integration with the API hook handlers [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) [16:22:08] awight: hey, can you take a look at https://github.com/wikimedia/ores/pull/284 ? [16:23:33] That one's my speed, even ;-) [16:24:42] Thank you! [16:24:46] going to deploy it now :D [16:26:32] wikimedia/ores#1138 (master - d319a37 : Amir Sarabadani): The build was broken. https://travis-ci.org/wikimedia/ores/builds/455578042 [16:34:00] I will be back soon [16:36:12] wikimedia/ores#1138 (master - d319a37 : Amir Sarabadani): The build was broken. https://travis-ci.org/wikimedia/ores/builds/455578042 [17:00:03] 10Scoring-platform-team, 10Wikilabels: Wikilabels needs manual reboot when DB connection is broken - https://phabricator.wikimedia.org/T209604 (10awight) [17:05:50] (03PS1) 10Ladsgroup: Bump ores to HEAD [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/473786 [17:06:31] awight: Also this: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ORES/+/473756 [17:06:35] :P [17:07:03] I'm pretty sure I pushed the second patchset [17:07:35] Great! Is that ASAP or can I take a minute? [17:07:49] Amir1: ^ [17:07:58] take your time [17:08:12] First, I need to know what's going on with the patchset [17:08:53] Amir1: Unrelated question, is wikilabels-02 connecting to a production DB server? Looks like pgsql.eqiad.wmnet [17:08:57] (03CR) 10Ladsgroup: [V: 032 C: 032] Bump ores to HEAD [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/473786 (owner: 10Ladsgroup) [17:09:41] pgsql.eqiad.wmnet is labsdb1004.eqiad.wmflabs [17:10:00] it's actually a labsdb, so no need to worry [17:10:01] Thanks! How did you know that... [17:10:37] After panicking :D [17:10:44] I think Madhu told me [17:12:14] hahaha sheesh. Yeah I could not find it in puppet so far [17:13:23] Now that I know what to look for, I can confirm it in site.pp, FWIW [17:14:37] weird, you'd think that a "threaded connection pool" would be able to recover from a server restart [17:16:29] All of this should be handled by pymysql [17:21:44] 10Scoring-platform-team (Current): Understand TeaHouse desires of Newcomer-Quality predictions - https://phabricator.wikimedia.org/T208367 (10notconfusing) - Jmo and I chatted about deploying session-based newcomer-quality ML model (NCM) for use in TeaHouse at CSCW2018 - Jmo sounded encouraged by early res... [17:21:51] ^ psycopg2 [17:22:00] Not pymysql [17:22:09] But I agree that it *should* recover. [17:22:27] I did some work to try to address that a while back but I failed to get it to work. [17:22:36] 10Scoring-platform-team (Current): Understand TeaHouse desires of Newcomer-Quality predictions - https://phabricator.wikimedia.org/T208367 (10notconfusing) p:05Triage>03Normal [17:24:41] I'm finding plenty of breadcrumbs, we can definitely make the reconnect work somehow, but it won't be transparent to our code. [17:25:01] 10Scoring-platform-team (Current): Create sample Newcomer quality predictions for TeaHouse hosts to sanity check - https://phabricator.wikimedia.org/T209607 (10notconfusing) [17:25:12] 10Scoring-platform-team (Current): Create sample Newcomer quality predictions for TeaHouse hosts to sanity check - https://phabricator.wikimedia.org/T209607 (10notconfusing) p:05Triage>03Normal [17:27:35] halfak: awight: Two things: I did a little bit of research on sentinel and it's a viable option to make our redis HA, I need akosiaris to give some input on how to proceed there but take a look at https://phabricator.wikimedia.org/T122676#4749717 [17:28:17] Second part: In order to reduce number of lots of requests people make to the service, I was thinking of making ores datasets in hadoop cluster and dumps of that table [17:28:42] 10Scoring-platform-team (Current): Post to enwiki TeaHouse about conducting AI-experiment - https://phabricator.wikimedia.org/T209608 (10notconfusing) [17:29:10] 10Scoring-platform-team (Current): Post to enwiki TeaHouse about conducting AI-experiment - https://phabricator.wikimedia.org/T209608 (10notconfusing) p:05Triage>03Normal [17:29:15] WMCS is working to make a hadoop/presto cluster in labs, it's definitely will be useful and we can move our stuff there once implemented to make it available for everyone [17:34:01] Amir1: HA-ifying Redis sounds awesome. Is the "redis cluster" package available yet, or is -sentinel definitely the way to go? [17:34:25] awight: it's already shipped with redis 2.6 or higher [17:34:33] we are on 3.2 [17:34:51] "it" = sentinel, I think [17:34:55] https://fnordig.de/2015/06/01/redis-sentinel-and-redis-cluster/ [17:34:59] +1 for getting ORES predictions in hadoop. [17:35:25] Been talking to joal about getting spark to run some of ORES models. [17:35:44] But I think that won't work in the short term and it won't work for many of our models. [17:36:07] So I think it would make more sense to (1) backfill and (2) consume the revision-scored stream. [17:36:21] ottomata is working on the revision-scored stream [17:36:26] That comes from ChangeProp [17:37:05] Is there a task for storing the scores in Hadoop? A researcher Claudia Muller-Birn has a use case that requires exactly that. [17:37:34] I think there is something but not sure if I can pick it up [17:37:40] I don't think so. I've just been working from the revision-scored task. [17:39:11] halfak: WRT running Spark, what if we were able to cache features in Hadoop as well? That should let us support all models, right? [17:41:20] 10ORES, 10Scoring-platform-team, 10revscoring, 10artificial-intelligence: Store ORES scores in Hadoop - https://phabricator.wikimedia.org/T209611 (10awight) [17:43:11] awight, not sure what you mean. [17:43:35] I could be misunderstanding what the obstacles are. [17:44:45] Is the issue that Spark only supports some algorithms? And would we still be running feature extraction from the main thread? [17:47:05] For the sake of deduplication, I would just listen to revision-score stream, put them into hadoop and develop a maintaince script to just push things there (for sake of backfill) [17:48:24] 10Scoring-platform-team, 10Wikilabels: Wikilabels needs manual reboot when DB connection is broken - https://phabricator.wikimedia.org/T209604 (10awight) The solution is probably to catch OperationalError and DatabaseError around cursor.execute calls. We would have to reconnect using a backoff, and refresh al... [17:51:36] That sounds like a sane approach... So I guess we would run the backfill job on a labs instance? Or would this be a low-level thing hitting production for a few weeks? [17:52:34] IMO the latter option is a good one. [17:52:58] I would run it stat1007 hitting prod for a couple of weeks (cronjob I guess) [17:53:02] We can score quite a lot over the course of a week without taking down ORES :) [17:53:06] I've done similar things for mediawiki before [17:53:17] Only problem is caching. We might want to allow a query param to skip the cache. [17:53:29] We don't want to fill our cache with a full scan of everything. [17:53:41] Then again, maybe that doesn't matter. [17:53:57] * halfak needs to run away. [17:58:57] +1 nice thought re: not caching [17:59:58] That's almost something that all large scans should turn on once the client's initial debugging is complete. [18:03:51] awight: Ping https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ORES/+/473756 :D [18:05:12] also https://github.com/wikimedia/ores/pull/287 [18:05:21] 10ORES, 10Scoring-platform-team, 10revscoring, 10artificial-intelligence: Store ORES scores in Hadoop - https://phabricator.wikimedia.org/T209611 (10awight) Notes from IRC: * Backfill old values by hitting production with a maintenance job. * Pipe new scores into ORES by reading from the changeprop stream.... [18:05:33] Amir1: Great, will do! [18:07:01] o/ [18:07:38] o/ [18:07:49] 10ORES, 10Scoring-platform-team, 10revscoring, 10artificial-intelligence: Store ORES scores in Hadoop - https://phabricator.wikimedia.org/T209611 (10awight) [18:09:50] btw. The only reason is codfw is super slow is that cache hit rate is pretty bad [18:10:03] it's very likely researchers [18:13:45] (03PS1) 10Ladsgroup: Bump ores to HEAD [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/473807 [18:14:24] (03CR) 10Ladsgroup: [V: 032 C: 032] Bump ores to HEAD [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/473807 (owner: 10Ladsgroup) [18:17:01] Interesting! Sounds like we should prioritize the new "nocache" flag. [18:18:07] (03CR) 10Awight: [C: 032] "Makes sense, good idea!" [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) (owner: 10Ladsgroup) [18:20:48] Oh another update, Using git lfs, I increased number of parallel connections from 3 to 9, the deployment time is now down to 14 minutes (it used to be 30-40 minutes) [18:21:04] <3 awesome work! [18:21:24] Too bad about 9 vs 10--did you hit a wall at 9? [18:21:31] oh. [18:21:38] we have 9 per DC, all good [18:21:42] 10ORES, 10Scoring-platform-team (Current), 10User-Ladsgroup: Investigate what is creating Redis transactions and whether it can be fixed - https://phabricator.wikimedia.org/T196889 (10Ladsgroup) 05Open>03Resolved a:03Ladsgroup [18:21:44] 10ORES, 10Scoring-platform-team, 10Patch-For-Review: Implement twemproxy for ORES production Redis - https://phabricator.wikimedia.org/T122676 (10Ladsgroup) [18:22:19] 10ORES, 10Scoring-platform-team (Current), 10User-Ladsgroup: Investigate what is creating Redis transactions and whether it can be fixed - https://phabricator.wikimedia.org/T196889 (10Ladsgroup) This is done, we are not fixing transcations. We'll work on sentinel [18:23:16] 10ORES, 10Scoring-platform-team: Implement sentinel for ORES production Redis - https://phabricator.wikimedia.org/T122676 (10Ladsgroup) [18:23:56] 18 would be better, though :) [18:24:21] (03Merged) 10jenkins-bot: Drop all of service integration with the API hook handlers [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) (owner: 10Ladsgroup) [18:24:24] I swear something's wrong with the entire premise, BTW. [18:24:26] I was thinking about it too. but it doesn't matter much [18:24:44] it would reduce it probably to ten minutes [18:25:22] 10ORES, 10Scoring-platform-team: Implement sentinel for ORES production Redis - https://phabricator.wikimedia.org/T122676 (10Ladsgroup) p:05Normal>03High [18:25:24] Most of these syncs should be instantaneous, cos the objects are available in local .git on each machine. We should only be transferring the updates. [18:25:53] 10Scoring-platform-team, 10Operations, 10Wikimedia-Incident: ORES overload incident, 2017-11-28 - https://phabricator.wikimedia.org/T181538 (10Ladsgroup) [18:25:56] 10Scoring-platform-team (Current), 10Operations, 10User-Ladsgroup, 10Wikimedia-Incident: Celery manager implodes horribly if Redis goes down - https://phabricator.wikimedia.org/T181632 (10Ladsgroup) 05Open>03Resolved [18:26:01] twentyafterfour says that the caching is working, so I can't imagine what the delays are. [18:26:09] ^ NICE [18:26:31] 10Scoring-platform-team (Current), 10Operations, 10User-Ladsgroup, 10Wikimedia-Incident: Investigate redis-cluster or other techniques for making Redis not a single point of failure. - https://phabricator.wikimedia.org/T181559 (10Ladsgroup) a:03Ladsgroup We decided to go with redis-sentinel [18:26:47] 10Scoring-platform-team, 10Operations, 10Wikimedia-Incident: ORES overload incident, 2017-11-28 - https://phabricator.wikimedia.org/T181538 (10Ladsgroup) [18:26:50] 10Scoring-platform-team (Current), 10Operations, 10User-Ladsgroup, 10Wikimedia-Incident: Investigate redis-cluster or other techniques for making Redis not a single point of failure. - https://phabricator.wikimedia.org/T181559 (10Ladsgroup) 05Open>03Resolved [18:27:20] (03CR) 10jenkins-bot: Drop all of service integration with the API hook handlers [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473756 (https://phabricator.wikimedia.org/T209459) (owner: 10Ladsgroup) [18:29:02] 10ORES, 10Scoring-platform-team (Current), 10User-Ladsgroup: Migrate ores celery configs to celery 4 - https://phabricator.wikimedia.org/T209587 (10Ladsgroup) a:03Ladsgroup [19:13:02] halAFK: how much longer are you in SF? [19:28:52] Until tomorrow afternoon [19:29:04] I forgot you were around! I'm going to add you to a dinner thread :) [20:00:55] I will be afk for dinner, I will be back soon [21:25:03] halAFK: hey, around? [21:27:55] Sort of. What's up? [21:27:58] Amir1, ^ [21:28:26] awight: one thing I am thinking of in the UI for editquality judgments is, rather than have two menus with two options (constructive/destructive; goodfaith/badfaith), we simplify it to one menu with three options that represent all of the matrix combinations except constructive/badfaith which I'm not convinced is a real thing. [21:28:44] Cool! [21:28:45] This would be strictly a UI convenience, and not a change to the underlying schema [21:29:22] so * damaging [and ill-intended], * non-damaging and bad faith, and * non-damaging and good faith [21:29:28] I like it. [21:29:41] ah no [21:29:44] what you said. [21:30:10] * nasty and damaging, * accidentally damaging, * harmless [21:30:26] Yes [21:30:37] Reducing the two "dimensions" to a single scale makes me feel warm and fuzzy [21:31:03] We could also pose it as a question. "Does this edit improve this page? Yes/No but they mean well/No and they don't mean well" [21:31:12] Bikesheds welcome [21:32:41] * harej tries to imagine the last case... "they contributed positively to the encyclopedia, but did so with malintent" [21:33:36] Just saying also, testing 3-choice vs. 2-checkbox would be a valid user testing question. [21:34:01] Yes. What's better: fewer things to fill out with slightly more complication, or two simple things to fill out? I honestly don't know the answer. Both seem valid for different reasons. [21:34:37] +1 conceptually they both make sense [21:35:04] With 2 buttons, they're very distinct concerns. With 3 choices, it's a single judgment. [21:40:58] 10JADE, 10Scoring-platform-team: Be able to flag judgments for additional participation - https://phabricator.wikimedia.org/T209649 (10Harej) [21:41:24] gtg for a bit [21:43:48] harej: could I chat with you about talking to the TeaHouse community? [21:44:22] what would you like to know? [21:48:07] back [21:48:14] harej: as you know, I've been making this newcomer-quality model based on ORES. One potential use for it, is to replace the current way that HostBot invites newcomers to the TeaHouse. Jmo and I chatted, currently HostBot invite 300 randomly selected newcomers every day (that have 5 edits and havent' been blocked). The newcomer-quality model could suggest the 300-best newcomers based on their ORES editing. We would like to conduct a [21:48:16] for about a month, to see if the AI approach could incraese retention. My question is, how should I pitch this to the TeaHouse community? Just post a plan on the talk page? [21:48:49] notconfusing: BTW do you have docs I can read about the project so far? [21:48:50] And if so, also how the narrative of the plan should go? [21:49:08] I'd also be interested in seeing docs, if you have any. [21:49:24] I would open by posting on the talk page about it, yeah [21:49:35] awight: I should make a page, eh? This is most of what I've got right now [21:49:37] https://en.wikipedia.org/wiki/Wikipedia:Labels/Newcomer_session_quality [21:49:42] I would describe how it works now and what change you're planning on making, and why. [21:49:49] Direct and clear communication is the key. [21:50:11] Should I be asking for "permission" to conduct the experiment? [21:50:27] Or rather just as if I was floating the idea? [21:50:45] notconfusing: Random thought--have you considered also analyzing Talk discussion quality to find the most constructively chatty editors? [21:51:12] the model at the moment doesn't distinguish between namespaces [21:51:15] i can add that [21:51:32] notconfusing: I think it's worth playing nice with them. I don't know that you officially have to ask for permission per se, and I wouldn't frame it like that. I would say "this is the proposal" and ask if there are problems with it. [21:52:42] Excellent idea. I was just wondering. In conversations do I have to disclaim that I'm working from foundation? As a contractor I never asked to make a "NAME (WMF)" account. [21:53:42] notconfusing: I think articlequality won't run on Talk pages. I'm imagining a new model anyway, so this would be something for a later iteration. [21:54:15] notconfusing: Declaring your WMF CoI could be done on your user page, if you aren't already... [21:54:25] gotcha [21:54:49] Sadly: https://ores.wikimedia.org/v3/scores/241703130 [21:55:09] https://ores.wikimedia.org/v3/scores/enwiki/241703130 [21:55:22] *ahem* thanks [21:56:25] which Wiki should i create a page about newcomer model on? Enwiki, meta, or mediawiki? I see ORES docs are on mediawiki, but the labelling campaign docs are on enwiki? [21:57:00] notconfusing, I think mw:ORES/Newcomerquality [21:57:05] ty [21:57:19] awight, and I were just talking about making mw:ORES/*quality better. [21:57:27] awight, ^ you agree on location? [21:57:28] Might also fit under meta:Research: for now? [21:58:25] For people in codfw area (west coast) is ores responds okay to you? [21:58:44] (check the responding node with the header) [21:58:54] Amir1, looks good. I'm connecting from SF [21:59:25] halAFK: Can you check the header and make sure it's codfw? [21:59:34] I'm getting eqiad responses [21:59:47] our cache hit rate there is 10% [21:59:53] while eqiad is 80 [22:00:39] That's just the type of traffic, right? [22:02:53] yeah but still can't say for sure [22:04:29] https://phabricator.wikimedia.org/T182222 might illuminate in the future... [22:08:13] notconfusing: I wanted to say earlier, this HostBot experiment seems like a smart and effective intervention. [22:08:42] I'm curious whether there are precedents for how to prioritize outreach in Wikimedia? [22:09:12] There's no canonical way, if you're wondering. [22:09:25] There are infinitely many forums for things. [22:09:39] Thanks awight! Also, that's what harej is for (adivising, not for doing). [22:09:42] Not looking for standards at this point, just want to know how it's done. [22:10:28] If it's a sitewide change you announce it on the wiki's village pump (preferably in their language) and get an entry in Tech News [22:11:15] If we can make uwsgi to spit the logs in a more structured way, logstash would easily handle it [22:11:19] More mature projects like the visual editor have a whole newsletter and all sorts of workflows worked out. [22:11:21] halAFK: I wish I could join this dinner thread too but need to stay home with my kid. I still want to research on contentious judgements and will contact you as soon as we get well. [22:11:28] but now, it's not possible [22:11:38] My Talk-scoring comment was coming from a consideration about the optimization function. editquality is mostly optimizing for perceived success in article content space, which has very different metrics e.g. citation count than talk space, or high engagement and inclusive discussions. [22:12:27] It would be interesting to optimize for sociality, specifically in the Teahouse context. We might be able to mine raw features for Teahouse edits made by its friendliest members, to learn what makes a good TH participant? [22:13:59] 10ORES, 10Scoring-platform-team (Current), 10Scap, 10User-Ladsgroup: ORES deployment finish "successfully" even when uwsgi and celery fail to successfully start up - https://phabricator.wikimedia.org/T170950 (10Ladsgroup) a:03Ladsgroup [22:14:22] 10JADE, 10Scoring-platform-team, 10Edit-Review-Improvements, 10Growth-Team: "Report error" button for ORES recent changes filter - https://phabricator.wikimedia.org/T209653 (10Harej) [22:14:43] 10JADE, 10Scoring-platform-team, 10Edit-Review-Improvements, 10Growth-Team: New watchlist filter: diffs without corresponding Judgment-space pages - https://phabricator.wikimedia.org/T209548 (10Harej) [22:14:48] 10ORES, 10Scoring-platform-team, 10monitoring, 10Wikimedia-Incident: Create Grafana graph to show number of ORES API requests per user-agent - https://phabricator.wikimedia.org/T182222 (10awight) Today this would have been useful for monitoring low cache hit rate in one data center. This still feels like... [22:15:33] I'm done for the day [22:15:51] see you tomorrow. I might not be able to participate at the sync meeting [22:15:57] o/ strong show today! [22:16:39] Thanks <3 [22:17:41] Just found this stale wishlist task, for a talk quality model similar to what I'm imagining: https://phabricator.wikimedia.org/T120431 [22:18:12] 10ORES, 10Scoring-platform-team, 10Community-Wishlist-Survey-2015: Machine-learning tool to reduce toxic talk page interactions - https://phabricator.wikimedia.org/T120431 (10awight) [22:21:19] 10ORES, 10Scoring-platform-team, 10Community-Wishlist-Survey-2015: Machine-learning tool to reduce toxic talk page interactions - https://phabricator.wikimedia.org/T120431 (10awight) I'm foster-parenting this task, since it's come up in our team a few times lately. Having models to estimate social behaviors... [22:23:13] 10ORES, 10Scoring-platform-team, 10Community-Wishlist-Survey-2015: Machine-learning tool to reduce toxic talk page interactions - https://phabricator.wikimedia.org/T120431 (10awight) This would probably derive some metrics from [[en:User:Ewitch51]]'s pending research on civility. [22:23:13] 10[1] 10https://en.wikipedia.org/wiki/User:Ewitch51 [22:24:48] Oh no Fajne! Forgot you were in the bay area too :( I hope your kid feels better soon. [23:02:21] (03CR) 10jenkins-bot: Localisation updates from https://translatewiki.net. [extensions/JADE] - 10https://gerrit.wikimedia.org/r/473897 (owner: 10L10n-bot) [23:09:03] (03CR) 10jenkins-bot: Localisation updates from https://translatewiki.net. [extensions/ORES] - 10https://gerrit.wikimedia.org/r/473900 (owner: 10L10n-bot) [23:31:09] 10ORES, 10Scoring-platform-team (Current), 10Growth-Team, 10MediaWiki-extensions-PageCuration, and 2 others: Merge articlequality and itemquality - https://phabricator.wikimedia.org/T206037 (10awight) I'd like to finish up this discussion. To summarize my current preference, I think that we should introdu... [23:45:40] halAFK: I hope you are taking care of your respiratory system and are not exerting yourself too much. The air is really, really bad today. [23:46:33] +1 ad-hoc filters like a bandana won't do anything [23:46:38] Na. Mostly sitting inside and working on workshop stuff. [23:46:57] The office has respirator masks if you want to swing by here on the way to dinner. [23:47:17] Hmm. I might do that. [23:47:30] You can experiment on yourself to see how much your tone has atrophied. [23:47:43] harej: (you could take a mask to dinner for AH!) [23:47:54] Ooh, I could. [23:48:16] I wouldn't recommend wearing it during dinner though. Might make eating harder. [23:48:25] * awight snatches that $4.50 from Muni's clutches [23:48:59] Could probably drink a beer with a mask on if you're really determined. [23:49:47] helmet/straw/mask integration commences [23:50:16] Hey, while the whole gang is sort of here, can we agree on T206037 ? [23:50:16] T206037: Merge articlequality and itemquality - https://phabricator.wikimedia.org/T206037 [23:50:26] My last comment summarizes what I'd like to do. [23:51:43] N95 masks are supposed to filter out particles 2.5 micrometers in size. Beer is bigger than that.