[14:21:32] New review: MaxSem; "Timo, you seem to misunderstand: currently, only what you call "classic" mode is present, so your po..." [mediawiki/extensions/MobileFrontend] (master); V: 0 C: 0; - https://gerrit.wikimedia.org/r/14419 [16:59:54] MaxSem: hello [16:59:59] hi [17:00:25] MaxSem: have you seen awjr? [17:00:29] nope [17:01:02] MaxSem: ha ha [17:01:09] awjr: hi [17:01:19] whatsup [17:01:32] sorry just realized it was 10 and i hadn't hopped on IRC yet [17:02:18] awjr: let's just do our meeting on IRC…. will that work? [17:02:27] preilly sure [17:03:15] MaxSem: you go first. [17:04:00] WLM! pending a sync with awjr just revamping the API [17:04:24] EOM [17:04:40] MaxSem: okay [17:04:46] awjr: your turn [17:06:01] friday was my 20% day, although i got a nice overview of zero and sibbling project redirection stuff. today, i was planning on observing some of the testing for zero, catching up with MaxSem about the API for WLM, and continuing to chip away at WLM bugs [17:06:11] no blockers. [17:07:27] awjr: okay sounds good [17:08:13] Friday was my 20% day I worked on embedding the V8 javascript engine into php https://github.com/preillyme/v8js [17:08:22] Today I'm working on Zero testing and fixes [17:08:41] awjr: this is the zero test for today [17:08:41] https://office.wikimedia.org/wiki/Partner_IP_Live_testing_schedule#Upcoming_testing_schedule [17:09:34] awjr: which was this change https://gerrit.wikimedia.org/r/#/c/14774/ [17:10:10] preilly cool [17:10:24] so is testing just enabling the ACL and waiting for feedback? [17:10:44] awjr: I've asked in #wikimedia-operations for the change to be approved and merged [17:10:56] awjr: yes that's pretty much it [17:11:27] awjr: I'm also playing around with simplifying the configuration process and architecture of the extension [17:11:43] preilly: where does testing feedback go? to some email address or something? [17:11:55] preilly: cool - what are changing? [17:11:59] awjr: no Dan Foy works directly with the carriers [17:12:18] awjr: I'm not sure what will be changing yet [17:12:24] awjr: I'm still noodling it [17:13:40] preilly word [17:17:54] awjr, ready to discuss WLM? [17:18:06] MaxSem: give me 5 minutes im just wrapping up an email [17:18:15] okay [17:22:01] MaxSem: ok im ready when you are [17:23:44] awjr, so... the current problem is FTS. I could either work with ops to deploy Solr or invent some poor man's FTS just like MySQL does it itself with fulltext indexes in 5.6 where the index is actually a table [17:24:16] MaxSem is lucene out of the question? [17:25:16] preilly said our existing lucene can't be used for generic searches [17:25:52] MaxSem have you talked with anyone in ops about solr yet? [17:26:20] was going to [17:26:37] cool [17:26:39] awjr, what's your opinion on all this? [17:26:56] that we should take the path of least resistence :p [17:27:14] implementing our own fulltext search is a bad idea, i think [17:27:41] have you done any testing with solr yet? [17:29:17] I've played with it a bit. basically, Solr is a search engine on top of Lucene [17:30:28] it will definitely handle out loads, however currently WMF lacks an engineer familiar with Solr/Lucene. ops are looking for a contractor, but no hires yet [17:31:13] i suspect our use-case is fairly basic for solr - is not having an engineer particularly familiar with solr/lucene going to be a problem for us if we use it? [17:33:12] for development - no, I can write everything MW-side (and a search schema for Solr) quickly. the problem here is deployment and fine-tuning [17:34:44] ok makes sense [17:34:55] who are you gonna talk to in ops about this stuff? [17:36:28] to the whole channel:) [17:38:16] heh ok [17:40:51] MaxSem aside from fulltext search what are the other outstanding issues? replication vs dumps/imports? [17:41:30] I think I've decided on this: replication would be PITA [17:42:16] yeah, from the email thread about the migration, it sounded to me like replication is not the best approach [17:43:02] what are you thinking instead? [17:44:16] have you considered using linterweb's tools? [17:44:38] Their dumps are generally more-current. [17:45:05] Amgine, ?? [17:45:20] They run live replications of all WMF projects. [17:45:48] Amgine, we were discussing replication in the opposite direction;) [17:46:08] oh. [17:46:09] awjr, so I'll have to invent an update script that downloads a DB dum from toolserver, imports it into a temp table and update the main table where needed [17:47:25] MaxSem what about having the bot write directly to a cluster db? [17:52:29] I don't think that it's permitted for security reasons. also, the bot writes to a zillion of auxillary tables that have to be joined together later - it's even greater PITA [17:53:11] blech [17:55:04] ok well a dump/update script should be pretty straigthforward [17:55:11] sounds like you're on the right track [17:55:22] let us know what ops has to say about solr [17:56:13] oh, one question i had about the monuments data - what db would the replicated data live in? [17:57:07] also, is the fulltext search used on the monuments_all table, or somewhere else? [18:00:30] I was thinking about deploying the API on Commons, however monuments_all is self-sufficient and can be both part of commonswiki or a separate database [18:01:36] 2) yes, we'll need only monuments_all and it needs a fulltext index [18:03:44] ok cool - you were probably already gonna do this when you talk to ops, but definitely be sure to make that clear when you're chatting with them [18:04:30] that will probably make folks less freaked out about a one-off solution, either with solr or enabling myisam and tuning for fulltext search in this particular case [18:05:44] okay [18:08:05] MySQL 5.6 supports FTS in InnoDB [18:08:17] yah [18:08:51] preilly, we'll just have to wait for it to make it into Ubuntu LTS [18:09:09] MaxSem: why is that? [18:09:23] MaxSem: we don't use the Ubuntu LTS version of MySQL in production [18:09:31] ops may be willing to do a one-off set up for this particular case [18:09:34] MaxSem: we use Facebook's MySQL [18:09:50] mmm, so many one offs [18:10:04] although i think ops has become less and less ok with doing one-offs over the last couple of years [18:10:16] if I were an op I would have killed myself as one-off:P [18:10:21] lol [18:40:50] who's working on the WLM app besides Yuvi? [18:41:48] i am, in theory [18:41:56] i've only just gotten started [18:41:58] and jon [18:42:14] jon owns the project [18:43:22] don't see him either [18:43:33] i think he's traveling to wikimania [18:44:33] oh [18:45:07] Jon is on a plane to D.C. ATM [21:23:42] hi, is it a known issue that the beta version doesn't seem to make a difference between whether the current article version is flagged or not (where the feature is enabled)? in the "normal" version, the latest flagged version is displayed, in the beta version, always the latest version is. (as far as i can see) [21:25:49] pill, I strongly doubt this - most likely, caching is borked [21:26:36] is it beta cookies bypassing cache, hmm [21:27:32] (just wanted to double-check but have no idea how to get back to the beta version ... hmm) [21:27:47] i tested it with http://de.wikipedia.org/w/index.php?title=Sigrid_Wylach&oldid=85504470&diff=cur [21:28:12] in the latest (unflagged) revision, the "Literatur" section has five entries, in the old one, four. [21:29:19] pill Special:MobileOptions [21:31:36] ok, yeah, it is as described. of course i don't know why :) [21:35:27] linSmith: hey, can you chat? [21:36:14] philinje sure [21:36:29] Skype? [21:36:32] yeah [21:36:39] one sec [21:37:06] greetings all [21:37:31] pchang__: how are things in the office ? [21:37:40] i saw the note about wikinews [21:37:45] should be easy enough to deal with [21:38:03] yeah, waiting for arthur to get back from lunch [21:38:10] very quiet here [21:48:29] shall i email anyone about it or is it just a normal cache effect? [21:49:25] looks like there might be a fix for the ICS can't-type-anything issue i saw with cordova 1.9 [21:49:26] whee [21:51:54] New review: Krinkle; "@MaxSem: I'm not misunderstanding it, I understand it very well. But whether what I referred to as "..." [mediawiki/extensions/MobileFrontend] (master); V: 0 C: 0; - https://gerrit.wikimedia.org/r/14419 [21:56:54] Krinkle, I'm still wondering if mw.loader can be decoupled from jQuery to work on more devices cause preserving the "old mode" would still result in versioned resource links for these devices and thus the need to purge varnish [21:58:30] MaxSem: Whether mw.loader uses it isn't relevant anymore. [21:58:31] What about all other modules? [21:58:32] Those use jquery as well. [21:59:03] The 2 or 3 jquery methods used in mediawiki.js can easily be implemented locally, Yes. But that would unneededly duplicate code from jQuery for no purpose. [21:59:54] MaxSem: What about dropping the classic mode and serving only "modern" and "no script" ? [22:00:05] we don't want to load all the stuff that gets loaded for desktop users - that would be disrespectful to users outside of the first world [22:00:22] I don't know what is currently covered in "old script" and about much percentage of traffic this is a bout. [22:00:52] Krinkle, no JS for old browsers is out of the question [22:01:05] MaxSem: Can you name an example module that is only useful on desktop and not on an iPad for example? [22:01:07] or iPhone. [22:01:30] every module that works only for desktop site [22:01:45] which currently includes most extensions, for example [22:01:55] examples please. there's too much vague "desktop only" talk in this whole atmosphere. [22:02:18] CategoryTree [22:02:25] Gadgets [22:02:28] MoodBar [22:02:41] ArticleFeedback? [22:04:04] DismissableSiteNotice [22:04:32] all seem perfectly possible for mobile as well [22:04:45] jquery sortable [22:04:47] collapsible [22:04:59] click tracking etc. [22:05:01] lots of modules [22:05:26] MoodBar will not work due to skin differences [22:05:44] same for DismissableSiteNotice [22:05:56] MoodBar can work just fine, all it needs is a few lines in css for skinStyles=>'mobile'=>'mobile.css' [22:06:10] that's exactly what that is for in resourceloader [22:06:20] so that stuff is only loaded for the relevant skin [22:07:02] so if that style of moodbar is specific for Vector, it can be moved in to vector property of the definition, and then it won;t be loaded on mobile [22:07:12] or better, add a mobile version, of course [22:07:15] and how MoodBar is supposed to work when there are no HTML elements for it to embed into? [22:07:46] Anyway, I'm not discussing MoodBar in particular now. [22:07:56] Made my point, waiting for reply. Right now dinners coming. [22:07:59] irc out, dc [22:08:00] the mobile skin is *completely* different [22:08:31] bleh [22:08:58] that was a classical desktop developer vs. mobile developer fight [22:09:27] btw - are we deploying today? [22:09:35] preilly, awjr_lunch ^^ [22:09:47] derp [22:10:11] MaxSem not that i'm aware of [22:10:14] * awjr checks the calendar [22:10:35] * MaxSem just received a calendar reminder [22:10:53] well it is the scheduled weekly update, but i have heard nothing about any updates needing deployment [22:15:53] were backing up on pull requests on WLM [22:15:55] https://github.com/wikimedia/WLMMobile/pulls [22:16:06] MaxSem: awjr: can eigther of you guys take a look at the queue ? [22:16:17] tfinc: im working on testing one of them already [22:16:23] awjr: k [22:16:55] tfinc: did you make it to dc already? [22:17:31] o/ [22:17:52] tfinc, I neither have the rights to merge this stuff nor experience with app dev'ment [22:19:26] tfinc, preilly, MaxSem, etc.: Any suggestions for the vumi-wikipedia User-Agent headere? [22:19:30] -e [22:19:51] We currently have: [22:19:52] 'User-Agent': 'Vumi HTTP Request', [22:20:00] awjr: yes. i'm working from my hotel room righ tnow [22:22:25] make it configurable to make it possible to point at exact Vumi installations? [22:25:20] MaxSem: We can do that. Configurable API URL would be good too. [22:25:25] * jerith cuts a ticket. [23:11:55] MaxSem: have we been approached by ru admins to blank ru.m ? [23:12:02] kul is telling me that we have [23:12:11] tfinc, haven't heard [23:12:22] if they do. what are our options ? [23:13:52] mmm, the mobile site is already unindexable by search engines. that somewhat helps [23:14:57] other than that - hack some stuff to return a banner message instead of wikitext [23:16:29] MaxSem: can admins blank it out without involving us ? [23:16:48] no [23:17:14] MaxSem: what could we do to enable that? [23:17:17] brion: yo! [23:17:18] heh, if my commit enabling mobile.js/mobile.css were merged... [23:17:34] tfinc: in the air dude [23:17:37] then they could do it themselves [23:17:39] i'm on a plane! on irc! [23:17:39] yay [23:17:45] liven' in the futureeeeee [23:17:47] how was the bkaclout handled on mobile when we did the US blackout? [23:18:03] there was no mobile blackout, just a banner [23:18:10] ah [23:19:25] I don't think we should attempt to black out mobile [23:19:28] brion: nice. can you do some in air code review ? were backed up on pull requests @ https://github.com/wikimedia/WLMMobile/pulls [23:19:36] tfinc: you read my mind dude [23:19:37] on it now [23:20:01] preilly: wether we do or not is a different discussion. i just want to know what are options are for doing it [23:20:37] preilly, especially with such late warning: the blackout vote began at 3pm on the 9th [23:21:00] brion, since you wrote MWSearch - can we use our existing search infrastructure for other text searches? [23:21:20] MaxSem: in theory, i suppose [23:21:52] oren bochman's doing some redesign on the server infrastructure but i think that's still in early stages [23:22:27] brion, like create a dummy "wiki", don't enable automatic updates to it, feed data into it manually with MWSearchUpdater? [23:22:38] MaxSem: *nod* sounds plausible :D [23:22:54] MWSearchUpdater may have bitrotted though [23:23:01] yeah, Oren's work is far from completion [23:23:35] * MaxSem is tempted to do a quick bitrot test in production XD [23:24:32] awjr: MaxSem how goes the WLM db migration ? [23:24:51] ^^^ [23:25:09] exploring non-MySQL search options [23:34:44] clustering looks nice [23:47:30] tfinc: ok only outstanding one now is https://github.com/wikimedia/WLMMobile/pulls is the progress one, i'll do more on that later [23:47:34] it's beverage time on flight [23:47:36] closing laptop :D [23:47:49] :) [23:52:40] preilly: when did MobileFrontend officially go live ? [23:58:25] tfinc: July 2011