[08:34:03] I don't know about the actual load time improvements due to ResourceLoader, but as far as I'm concerned, the *perceived* load times have worsened. [19:03:04] RoanKattouw: Yesterday mentioned something off-beat about a weekly meeting. Is this someting can/should attend ? [19:03:33] Once we get you set up with a contract and a phone ext and all that jazz, I guess so [19:03:53] k [19:15:21] Roan? [19:15:37] Yes [19:16:05] i gotta hypothetical for you. [19:16:22] say i wanted to find out the rates of people who bail from the wiki editor. [19:16:41] Rate as in what divided by what? [19:16:50] The number of people who chose to turn of wikieditor is easy to find [19:17:05] one way we could do that is to create editor tokens. you click "edit", we generate a unique token, store it in the page and database (db has timestamp, url, etc.). when you click "save", we update that token as "used". [19:17:17] *off [19:17:22] then, we can check to see how many people successfully save versus how many bail without saving. [19:17:25] Oh, that [19:17:38] Erik Z has some stats on that, but that's based on analyzing sampled logs [19:17:55] we are finding that is not sufficient. [19:18:00] Basically he's just counting the number of requests for the edit page vs. the number of submits [19:18:07] But it sounds like you want more exact data [19:18:17] right. [19:18:34] You could use tokens, but counters sound simpler [19:18:48] Unless you actually want to use the timestamp and URL data [19:18:54] have to be tokens, so that we can slice the data on the user type. [19:19:03] how many newbies fail, how many experts fail, etc. [19:19:51] so, like, the token table would have to be something like: tokenID, userName, timestamp, completed [19:19:52] So you need to either link to user IDs or anonymously collect more dta [19:20:40] timestamp is useful so we can run another point: how long to save. but also: we can assume that if you haven't saved after an hour, you aren't going to. [19:20:41] Sounds sane [19:20:57] how difficult to get something like that? [19:20:59] tokenID can be auto_increment instead of a hash in that case [19:21:20] Shouldn't be very hard [19:21:23] Couple hours of work I guess [19:21:25] right. since it's unique per grab, we can just increment. [19:21:30] that's what i figured. [19:21:35] any obvious performance problems? [19:21:48] Yeah you can just insert the row first, then ask the DB "what ID did that end up getting" [19:22:04] None other than that we're talking about a massive shitload of data [19:22:22] we can run purges on it after rollups. [19:22:54] Hmm [19:22:59] there might be value in being able to historically follow a user's fail rate but i doubt that we'd learn anything useful over longer than a couple weeks. [19:23:07] I wonder whether we currently need to open a master conn for an edit page view [19:23:51] you would, wouldn't you? for anonymous? because you need the most recent version, always, and can't serve a cached copy. [19:24:07] Hmm, I guess so [19:24:57] not that i know exactly what i'm talking about here. [19:25:10] Pff [19:25:14] getcontent -> loadcontent -> fetchcontent [19:25:20] Thank you, authors of Article.php [19:25:26] would this be done as an extension or directly in the editor? [19:26:41] Hmm text fetches are slave first, then master [19:26:47] Wonder where the revision row comes from though [19:27:02] Probably doable as an extension, there's lots of hooks in that neighborhood [19:27:19] yeah, i can't see this being anything except WMF specific. [19:27:39] so, what this speaks to, so that we have some context: [19:27:50] we have a big-ass "editor trends" study. and that tells us a lot. [19:27:58] but now i'm looking at "editor lifecycle" [19:28:37] I'll just take the easy way out and find out with shotgun debugging [19:28:39] Right [19:28:39] and being able to see if people bail, when they bail, and where they bail will be . . . invaluable. [19:29:21] i'm gonna write something up on this. when i'm done, can you look it over and give me a sanity check on it? [19:29:33] people have been saying OWA can do this but I don't think so. [19:29:57] i have another thing i want to start tracking, too, which is "edits that are reverts". but we can do that with checksums. [19:31:01] Aha [19:31:13] Editing a page checks the rev timestamp, which opens a master conn, so we're good there [19:31:31] Sure, I can read your writings and comment [19:33:03] excellent. may be a couple days. [19:33:26] Sure [19:39:59] the completed column would then be a boolean value ? Or perhaps the rev_id ? [19:40:11] in the latter case you instantly have namespace, title and save time [19:40:17] save-timestamp [19:40:41] ie. rev_time - mystats.timestamp = editing time [19:45:01] yeah, revid is probably better. [19:45:20] because then we can later track it against reversion rates. [20:36:28] OK there's something very weird and hypocritical about this [20:36:30] "console is not defined" errors in the Firebug console [20:38:29] :D [20:59:45] RoanKattouw: http://www.mediawiki.org/w/index.php?limit=100&path=%2Ftrunk%2Fphase3%2Fresources%2F&title=Special%3ACode%2FMediaWiki%2Fpath [20:59:52] http://www.mediawiki.org/wiki/Special:Code/MediaWiki/82496 [21:00:34] RoanKattouw: I noticed you (intentionally?) use mediaWiki full name in recent commits. [21:00:35] Meh [21:00:43] Oh [21:01:03] I should really look at mw.js to see if that's necessary [21:01:24] except in mw.js itself, mw is available, everywhere [21:01:33] Looks like it, yes [21:01:39] So I guess I could just start using mw everywhere [21:01:44] mediaWiki = {}; window.mw = mediaWiki; [21:01:51] jep [21:03:57] There's a fair bit of optimzalization and minification" [21:04:11] There's a fair bit of optimzalization and "minification" possible by using the new prototypes and aliases in-core and extensions in SVN. [21:04:27] $.escapeRE, wikiGetLink, etc. [21:13:27] RoanKattouw: Hm.. can you see if there's any mismatches between aliases for root directories on-wiki and for bits.wm.o ? Seems for example /skins-1.17 is not available from a wiki's root [21:13:43] Does it matter? [21:13:53] 404 http://commons.wikimedia.org/skins-1.17/monobook/external.png [21:13:53] Hmm I guess it does for csshover.htc [21:13:57] indeed [21:14:07] and site css that uses relative paths [21:14:35] rare, but occurs (just saw it somewhere) [21:14:39] Fixed [21:17:51] per http://www.mediawiki.org/wiki/Berlin_Hackathon_2011 it looks like the Berlin hackfest will probably be late May, like 20-23 May [21:17:58] robla: ^ [21:25:11] whee: http://www.mediawiki.org/wiki/ControversialContent [21:25:19] pls to look over and give comments. [21:25:22] thx sumanah [21:26:07] CamelCase [21:26:17] BringsBackNostalgiaFromTheUseModEra [21:27:31] RoanKattouw: did you actually witness that era? [21:27:54] No [21:28:03] But that doesn't mean I can't make fun of people [21:30:22] eh. i've got many years of java. it's a habit. [21:31:48] The second image in that document deserves so many kudos [21:32:03] Use of Pokemon as an example, use of "Gladiatorial Combat" [21:33:08] "Ideally, an administrative interface could be developed to allow wikis to easily define a set of auto-hidden categories. However, in a minimal sense, these categories can be defined in the wiki's LocalSettings.php file as an array ($wgAutoHiddenContentCategories)." [21:33:20] jorm: If, by your own admission, there is a mother category for all controversial content flagging categories, why not just configure that one? [21:33:52] In a $wg var, or maybe even in a message so it can be changed through an MW: page [21:36:18] jorm: Also, what's that login dialog doing there? I want AJAX login as much as the next guy but I don't think shoving it into this feature is particularly appropriate [22:35:20] Nikerabbit: when I paste the text from [[Project:About/ksh]] into another page and preview it, I don't see the error [22:35:24] do you get an exception if you preview it on the same page? [22:35:24] or is it just on page view? [22:35:24] Hmm. A user preference to set debug=true would be very helpful when I can't just put a $wg to true RoanKattouw (eg. a wmf wiki) :P [22:35:25] ie. when submitting forms or whatever [22:35:28] it would make all the bugs go away though [22:35:28] you can't debug when there's no bugs [22:35:32] TimStarling did you create a user on translatewiki.net ? [22:35:32] yes [22:35:32] Ok I give you translator rights [22:35:32] TimStarling: I get the exception just by accessing the url http://translatewiki.net/wiki/Project:About/ksh [22:35:32] so do I [22:35:33] but the backtrace doesn't help [22:35:33] so I want to know what text or code causes it [22:35:35] the annoying thing is that this works: http://translatewiki.net/sw/Project:About/ksh [22:35:40] hmm, so it can't be previewed... [22:35:42] TimStarling: okay, you can ignore that one [22:35:43] I removed disabled few livehacks and it works now [22:35:43] ok, thanks