[05:01:56] hi Ryan_Lane [05:02:03] howdy [05:02:29] food is done. back in a bit :) [05:04:34] okay [05:04:52] just wondering what happened to the cldr thing? [15:55:18] anyone here? [15:57:02] mornin' howief [15:57:08] hey there [15:57:16] i just sent you a msg on gchat [15:57:35] are we still on for deployment? [15:57:37] ah yes [15:57:39] afaik [15:57:50] i haven't heard from roan yet [15:57:53] ok [15:58:26] or trevor for that matter [15:58:33] Morning folks [15:58:39] there he is! [15:58:40] speak of the devil [15:58:41] Sorry for being slightly late [15:58:44] heya RoanKattouw [15:58:45] no worries [15:59:57] RoanKattouw: thanks for those last additional fixes [16:00:04] Sure [16:00:14] The thing with the special page class is it's just generally bad practice [16:00:22] Special page classes are for UI/display purposes [16:00:24] deployment of AFT 3? [16:00:37] yes guillom [16:00:38] what would be the best practice way to handle that scenario? [16:00:41] Putting reusable logic in such classes is usually a bad idea [16:00:44] Especially if non-static [16:00:46] copy [16:00:47] ok [16:00:55] howief, do you have a blog announcement ready? [16:01:00] is the latest on prototype? [16:01:01] Well in this case it's just one function [16:01:09] So you just say "meh", make the function static, and live with it [16:01:12] howief: Let me update prototype [16:01:19] guillom: we haven't finalized the blog post [16:01:27] ok, just asking [16:01:30] but i figure we can just use what erik sent to wiki-en [16:01:38] thanks RoanKattouw [16:04:19] Alright, prototype updated [16:05:04] great [16:05:30] *RoanKattouw waves at TrevorParscal [16:05:34] howdy [16:05:41] whattup TrevorParscal [16:05:43] I just merged that $.appear rev you asked for, and updated prototype [16:06:00] cool [16:06:00] so has the dashboard page not been updated? [16:06:00] thanks [16:06:13] i'm still seeing the old copy [16:06:37] http://prototype.wikimedia.org/release-en/ArticleFeedbackTest is working [16:06:59] http://prototype.wikimedia.org/release-en/Special:ArticleFeedback - i see 2 tables... [16:07:21] no colors - i guess the default sort never got... sorted [16:07:56] nor is any of the additional messaging present [16:08:05] would it possible to make the changes prior to deployment? [16:08:47] quick poll [16:09:15] what do people think of the new messaging in the green, which appears upon the first click of a star? [16:09:17] Depends on their magnitude [16:09:25] would that make you click on "submit ratings"? [16:10:34] i'd like RoanKattouw guillom awjr and anyone else on the channel to respond [16:10:48] (since TrevorParscal and i were the ones working on it) [16:11:10] I'll just say that such design decisions should be considered more than -18 minutes before scheduled deployment [16:11:11] howief: i like it, it definitely grabbed my attn [16:11:21] agreed with RoanKattow [16:11:24] *RoanKattouw [16:11:30] RoanKattouw: i wasn't able to see this on prototype until now [16:11:38] hmm [16:11:42] but otherwise, i agree with you [16:11:43] howief: when I used a non-green color, it looked like you did something wrong [16:11:44] Oh, cool [16:11:46] like an error or warning [16:11:50] Yeah that looks useful [16:11:54] TrevorParscal: yes, that was very clever :) [16:11:54] howief, I think it's less ideal than being saved automagically, but it's better than nothing. [16:11:59] And good call on the green, Trevo [16:12:01] r [16:12:08] plus, i tend to not be the best person to ask about these kinds of things since my preferred UI is the terminal... [16:12:12] hehe [16:12:21] I'm with guillom , it's just a patch for bad UX, the fundamental issue isn't fixed [16:12:21] awjr: would it have compelled you to click on "submit"? [16:12:24] And +1 on awjr [16:12:31] howief: yes [16:12:32] ok [16:12:37] let's roll with it then [16:12:44] and i do agree with guillom [16:12:56] this was just a last minute thing we discovered from the UX testing erik did [16:12:57] Oh, crap [16:13:03] (topic is ArticleFeedback?) [16:13:04] I did not look at the Squid stuff from Trevor at all [16:13:18] brion reviewed it a bit [16:13:23] and I responded with a patch [16:14:24] It's incomplete [16:14:36] The URL, both in JS and in PHP, needs smaxage=somenumber added [16:14:42] Otherwise Squid won't cache it at all [16:14:49] RoanKattouw: yes, I know [16:14:57] that's the area I was hoping you could help... [16:15:16] should be trivial [16:15:25] Add &smaxage= an appropriate number to the query strings on both ends [16:15:37] Said number should probably be configurable [16:15:43] what about maxage? [16:15:51] That's for client-side caching [16:15:54] yeah [16:16:00] Shouldn't be set too high [16:16:03] i'm asking, do we need that in the URL too? [16:16:05] We're using on-demand purging here [16:16:05] both right? [16:16:09] yup [16:16:13] Not if you don't want that [16:16:25] For on-demand purging of articles we use maxage=0 and smaxage=25920000 [16:16:30] ok [16:16:38] so, we always ask squid [16:16:51] and squid gets purged when someone rates [16:17:05] Ah, sans one zero [16:17:06] Yes [16:17:34] Can you implement this like right now so we can push it along with the rest? [16:17:44] morning all [16:17:48] yes [16:17:53] should have a patch in 1 min [16:17:55] sup brion [16:18:11] *TrevorParscal is also doing some digital packing so he has enough data to survive his trip [16:18:30] file transfers.... even over fw800 are slow [16:19:57] ugh, eclipse is also slow [16:20:31] it crashed... arg! [16:21:10] *RoanKattouw hands TrevorParscal a vim [16:21:16] hahaha [16:21:16] srsly [16:29:34] RoanKattouw are you discussing getting UploadWizard live ? [16:29:49] No, this is the next version of ArticleFeedback [16:29:52] v3 or v3.5 or whatever [16:29:55] ah [16:30:01] v4 [16:30:04] is UW on the roll [16:30:20] going by the bucket version [16:31:20] RoanKattouw: http://www.mediawiki.org/wiki/Special:Code/MediaWiki/87730 [16:31:46] GerardM-: At 9pm our time [16:33:02] RoanKattouw, TrevorParscal: smaxage should = 2592000, not 25920000, correct? [16:33:13] Yes [16:33:19] RoanKattouw Ah, sans one zero [16:33:46] <^demon> I love that we've hit v4 of UW and it's taken !10 years & we don't even have MW 2.0 ;-) [16:34:04] <^demon> ~10 years [16:35:13] Fixed in r87732 [16:35:32] Alright, let's go to test [16:36:36] howief: are we ok to launch the dashboard without the additional messaging? [16:38:47] hmmmm [16:38:52] i'd really rather not [16:39:09] but objectively, it's not a showstopper [16:39:37] sorry for my ignorance, but can the text of a special be edited while in production? [16:39:41] (special page) [16:40:10] i can easily and quickly add in the messaging but i dunno how you want it to look and am not the best person to have handle styling [16:40:25] Can we do that, at least for the "Experimental" bit? [16:40:38] I think a line of text above the title of the first table is good enough [16:40:46] and to answer your question, no, at least not like a regular wiki page [16:40:52] Lies [16:40:54] You can edit it [16:41:05] what?! [16:41:09] At [[MediaWiki:Articlefeedback-blah-blahblah]] [16:41:52] if that's the case, can someone edit post-deployment? [16:42:03] just to reduce the number of moving parts right now [16:42:11] We can also just edit the source and redeploy it [16:42:15] Deploying msg changes is trivial [16:42:38] whatever you think would be easiest [16:43:40] yeah, if you just want a line of text, no big deal [16:43:54] cool [16:44:02] RoanKattouw: im totally not seeing how you edit copy on a special page through the mw ui [16:44:51] You edit the i18n message(s) [16:44:58] ... [16:45:00] well yes [16:45:03] Through the MediaWiki: namespace [16:45:18] maybe i just dont have permissions to do that [16:45:39] You need to be a sysop to edit MediaWiki: pages [16:45:52] ah right [16:46:44] Wow, is it me or is AFT insanely slow on testwiki ? [16:46:55] Takes almost a minute for it to pop up. [16:50:01] Yeah it was slow the first time [16:50:10] Probably because its doing XHR, twice, and it's testwiki [16:50:26] Also, TrevorParscal , we overlooked something [16:50:31] Dashboard ? [16:50:34] There should be no smaxage for logged-in users [16:50:44] That'll just pollute Squid to bits [16:51:30] oh [16:51:30] Dashboard is part of this deployment, right ? [16:51:45] Wait, I *am* an anon and jQuery is doing the anti-cache thing [16:51:48] That's also bad [16:51:52] Krinkle: Yes [16:51:55] >_< [16:51:55] k [16:51:59] But... I guess it's disabled in the config [16:52:02] It's currently disa~^ [16:52:12] howief: i updated the special page to have the copy on it [16:52:16] great [16:52:18] it's live on prototype [16:52:32] Oh, cookie [16:52:33] it's ugly but it's there [16:52:35] looks great [16:52:45] :p [16:52:51] can we add another line break after the message? [16:53:02] RoanKattouw: patch in [16:53:12] you added the description of the averages too -- awesome [16:53:13] hey howief - morning folks, dario here [16:53:20] hey DarTar [16:53:30] hi DarTar [16:53:32] Dude, great minds [16:53:35] I wrote the exact same patch [16:53:37] hi awjr [16:55:06] I don't know if this has already been discussed but there was another comment yesterday regarding backward-compatibility with classic: http://www.mediawiki.org/wiki/Talk:Article_feedback#Broken [16:55:55] yes, i sent a note to trevor and timo about this [16:56:22] i think we should try to suppress the tool for classic if possible [16:56:41] RoanKattouw: are you on the hook for the UW deploy in an hour? [16:57:01] but i'm not sure that should block deployment [16:57:06] Yes, but we can push it back [16:57:08] Neil knows that [16:57:14] k [16:57:24] awjr: is carrie in the office right now? [16:57:31] TrevorParscal: she just walked in [16:57:42] i need you to get something in my desk, and give it to her, to give to me [16:57:48] k [16:58:09] any hints as to what it is? [16:58:22] Weapons. [16:58:25] in the bottom or 2nd to bottom drawer, there will be a box that looks like this http://www.myservice.com/blog/wp-content/uploads/2009/06/Picture-24-300x291.png [16:58:32] awjr, pick something at random [16:59:03] Bah [16:59:06] *RoanKattouw glares at TrevorParscal [16:59:09] cache: userrating is wrong [16:59:15] Should be cache: !userrating [16:59:16] also, a half eaten piece of gum, paperclip and a small box of mint flavored dental floss... that's critical for when I have to create a hang glider and escape the russians [16:59:29] arg! [16:59:38] Committing fix [17:00:31] just beat you [17:00:32] so on IE8 [17:00:36] Dammit [17:00:50] *TrevorParscal is a bug sniper [17:01:04] I submitted a rating, clicked knowledgeable, submitted an email, then created an account. when I went back to the page I rated, my ratings didn't appear [17:01:20] Were you an anon with no prior ratings? [17:01:32] i was previously anon [17:01:33] The logic for "cache or don't cache" was backwards [17:01:42] not sure if i had prior ratings [17:01:46] ok [17:02:19] TrevorParscal: found it, i'll see if i can convince carrie to bring it for ya [17:02:33] howief: You are testing on test.wikipedia.org , right? [17:02:39] yes [17:03:01] thanks man! [17:03:20] without that, I will be visiting an apple store in the netherlands [17:03:33] where you pay with cookies... [17:03:38] precious, delicious cookies [17:03:56] Oooh [17:04:00] I find it works better with colourful money that isn't all green. [17:04:03] And the AFT module doesn't specify a caching mode [17:04:18] So the caching isn't working [17:04:20] *RoanKattouw fixes [17:05:23] TrevorParscal: i believe she's emptying the box and just bringing the contents [17:05:35] Think of the apple box! [17:05:40] heh [17:05:49] is the daily meeting still on for today ? [17:05:51] also, isn't only the eu adapter useful? [17:06:13] Krinkle: i think we're postponing on account of the deployment [17:06:31] awjr: I only need the 2 adapters [17:06:40] actually [17:06:53] the euro plug tips, right? [17:07:03] other query from the AFT talk page, [17:07:12] so far we've been tracking the list of AFT-powered articles via the hidden category, we are not going to display a public list of all articles with AFT enabled from now on, are we? [17:07:28] could you ask her to just bring me the 2 ones that look sorta like this: http://m2.sourcingmap.com/smapimg/en/n/09a/apple-adapter-30360n.jpg [17:07:34] DarTar: No [17:07:51] k [17:09:20] krinkle: we're not planning to do a standup today since deployment is underway :-) [17:09:25] k [17:11:25] TrevorParscal: message relayed [17:14:05] cool beans [17:14:48] OK, I am reasonably convinced Squid caching is working now [17:16:20] huzzah [17:16:35] RoanKattouw: awesome [17:16:58] howief: Any objections to taking what's on test now to production? [17:17:01] howief: are you satisfied with how things look on prototype? [17:17:03] jinx [17:17:25] can we each do a quick run through on some browsers? [17:17:27] i'll do ie8 [17:17:35] Why is #catlinks on top of the page in useskin=standard/classic ? [17:19:49] howie: ok [17:20:43] my daughter is crushed that I'm not brining her with me on the airplane [17:20:45] :( [17:20:51] haha [17:20:58] Well she wants to go on the plane /now/ [17:21:00] TrevorParscal: aaw [17:21:05] looks good on IE8 [17:21:15] *Reedy blinks [17:21:16] chick, modern, monobook, and vector work [17:21:16] If and when you do bring her on a plane some time, you should remind her of this when she gets bored after an hour :) [17:21:23] cologneblue seems to just not show it [17:21:42] TrevorParscal: It's all dependant on #catlinks [17:22:00] FF 3.6 and 4.0 look good on os x [17:22:01] of course - I'm just saying, the ones that I see it working [17:22:09] $( '
' ).articleFeedback( config ).insertBefore( '#catlinks' ); [17:22:17] how about we use mw.util.$content ? [17:22:21] great [17:22:31] i'll check chrome on xp [17:22:32] Hm.. that would put it after catlinks though [17:22:38] can someone check safari on osx? [17:22:48] Krinkle: if you use mw.util.$content it will move it below the catlinks [17:22:52] you should fallback to that [17:22:54] I'm not sure browsers matter related to the positioning/absence of AFT in skins. [17:23:07] use catlinks if you can, otherwise use mw.util.$content [17:23:14] k [17:23:23] are you patching that now? [17:23:30] I need to start getting on my way [17:23:46] have to grab something at a store on my way to the airport [17:24:24] howief: safari looks good [17:24:34] yeah, I'm fixing the fallback now Trevor [17:24:34] awesome [17:24:43] firefox and chrome look good on xp [17:24:56] dashboard page isn't showing though [17:25:14] "Page has been disabled" message [17:25:22] howief: where are you testing? [17:25:26] Hm.. TrevorParscal if ( $( '#catlinks' ).length ) { isn't a reliable way to check it. Since that is present in useskin=standard as well, it's just that in standard #catlinks is on top instead of on the bottom [17:25:27] test.wikipedia.org [17:25:43] try prototype [17:25:50] http://pastebin.com/u4X1nWTX [17:25:53] i dont think code's been updated on test yet [17:25:55] Krinkle: you can special case that skin? [17:26:16] check if #catlinks is at the top somehow? [17:26:40] RoanKattouw: do you have enough of a grasp on how to ramp up AFT on the cluster to do w/o TrevorParscal around? [17:26:44] what's the default sort? [17:26:46] Yes [17:26:49] great [17:26:54] is it average, lowest to highest? [17:27:19] if catlinks directly following subtitle? [17:27:54] howief: i believe so [17:28:03] ok [17:28:15] we'll need to change that soon [17:28:20] but everything else looks good [17:28:33] let's go fowrard [17:28:48] okay, the pasttebin'ed code fixes it for skin=cologneblue (previously: not inserted in DOM due to no #catlinks; now: Appended to mw.util.$content) [17:29:12] What pastebinned code? [17:29:25] http://pastebin.com/u4X1nWTX [17:29:39] Oh. My. God. [17:29:46] The BZ autocomplete thing uses a JSON RPC thing [17:30:00] Apparently it POSTS a JSON object explaining what it wants or something [17:30:38] ok, I am switching to irc on my phone [17:35:03] RoanKattouw: what's the status, are we ready to start pushing things to the clusteR? [17:35:11] I am [17:35:16] Waiting on green light from Howie [17:35:24] howief gave the go-ahead [17:35:29] Oh, right [17:35:32] I missed that [17:35:40] I'm sorry, I'll go right ahead and push [17:36:10] k, lmk if there's anything i can do to help [17:36:32] are we pushing to 100k all at once? [17:36:44] i believe the plan was to slowly ramp up [17:36:47] k [17:37:03] but i dont think the plan was much more detailed that than [17:37:06] er, than that [17:37:12] ok [17:37:13] We are going to ramp up [17:37:20] We've spent the past hour and a half just doing the software change [17:37:33] sweet [17:37:49] if i'm calculating correct, the target is about 2.9%? [17:37:58] I.... don't know? [17:38:03] I guess [17:38:16] 100,000/3,500,000 [17:38:46] Well it's all main namespace pages, not just 'articles' [17:39:03] ? [17:39:08] We can safely exclude redirects [17:39:19] ah [17:39:23] howief: The number of 'articles' in the statistics page counts according to more stringent rules [17:39:32] that's right [17:39:33] good catch [17:39:59] Alright, sync done [17:40:03] New AFT version is live on production [17:40:28] Toolserver query for # of eligible pages still running :) [17:41:17] 3663178 [17:41:21] So that's not far off [17:41:28] k [17:41:33] let's go with 3% then [17:41:34] 2.7% [17:41:37] Yeah [17:41:50] That'll be ~110k, good enough [17:41:54] cool [17:42:51] Alright, so how many articles are we currently on/ [17:43:20] 3k i believe [17:43:24] yeah [17:44:01] 4,203 [17:44:26] Alright, so let's start with... [17:44:39] adding 22k, one-fifth of our target [17:45:07] drum roll. . . . [17:45:29] Or 18k, rather, that's 0.05% [17:45:45] do we have someone from ops here? [17:45:46] neilk_: Good morning. Our UW deployment is gonna have to wait a little while, probably until noon [17:45:59] RoanKattouw: okay [17:46:01] howief: mark is around and talking at the very least [17:46:09] As is Ryan_Lane [17:46:10] k [17:46:16] mark , Ryan_Lane : We're doing that ArticleFeedback ramp-up thing now [17:46:32] ok [17:46:39] i'm in meetings but can drop out if necessary [17:46:44] RoanKattouw: ok [17:47:10] TrevorParscalMob: Does $wgArticleFeedbackLotteryOdds still work? [17:47:16] Or did the way to specify that change too? [17:48:39] Ahm, not 0.05%, 0.5% [17:49:17] Alright, here we go, to 0.5% of all articles [17:50:35] Now live on 0.5% [17:50:40] w00t [17:50:50] nice! [17:50:57] Well [17:50:58] wheee [17:51:00] In 5-10 minutes [17:51:08] As the RL cache timeout runs out [17:51:43] So now we just sit around and wait for a bit while I figure out which articles I just deployed AFT to [17:51:43] RoanKattouw: as a side note, i see the dashabord is not enabled yet? [17:52:04] No idea [17:52:07] Is there a config var for it? [17:52:11] yes [17:52:14] If so, go ahead and set it yourself [17:52:16] i believe so [17:52:16] k [17:52:21] did you already run update.php? [17:52:26] Oops no [17:52:29] Didn't do schema changes [17:52:31] k [17:52:37] (Don't do that, I'll do it) [17:52:38] shall i do that? [17:52:39] k [17:52:44] i'll fix the config [17:52:44] Well, you can [17:52:50] As long as you don't literally run update.php [17:53:17] perhaps you ought to do it unless you've got a few extra mins to walk me through schema changes on the cluster [17:54:04] I'm doing it now [17:58:41] <^demon> Running update.php would be an awful awful idea :p [17:59:10] Yes [17:59:12] Schema change is done BTW [18:02:16] OK everything looks fine, let's ramp up further [18:02:21] I'm gonna go to 1% [18:02:33] RoanKattouw: im seeing uncommited changes to CommonSettings updating the version for AFT tracking/options [18:02:51] great [18:03:00] Oh, whoops [18:03:02] Let me sync those [18:03:05] from v3 -> v4 - is it ok to commit those? [18:03:24] ok - i just added config option to enable dashboard on enwiki as well, feel free to sync [18:03:35] Alright [18:04:40] RoanKattouw: what way are you ramping this up? [18:04:52] number of articles, then amount of clicktracking? [18:05:02] clicktracking is disabled currently, right? [18:05:10] # of articles is being ramped up [18:05:19] I'm not ramping down clicktracking yet but I should [18:05:31] turn it off completely [18:05:43] I'd like to know what it looks like when just AF is enabled [18:06:30] (this is what was discussed in the meeting last week, btw) [18:07:24] Alright, disabling [18:07:40] There is a 5-10 min latency for both changes (the bump to 1% and disabling click tracking) to take effect [18:08:40] click tracking was set to 50% previously, right? [18:08:50] that sounds fairly suicidal for a large number of articles :) [18:08:50] Yes [18:09:15] Yeah I was gonna ramp it down proportionally [18:09:30] better to ramp it up while looking at load [18:09:47] I meant that, as the # of pages ramps up, I was gonna ramp the tracking percentage down [18:09:55] yeah [18:10:05] 4k pages at 50% should produce the same volume as 100k pages at 2% [18:10:46] should [18:11:04] Anyway, I don't see anything happening on Ganglia [18:11:31] In fact, I haven't seen any of my AFT changes make any difference [18:12:20] TrevorParscal: dashboard enabled on enwp but not seeing messaging for table header labels (eg articlefeedback-rating-trustworthy)? [18:12:34] OK, now what [18:12:38] No AFT on [[Agatha Christie]] [18:12:57] Oh, wait [18:13:07] i just revisted a page that i had previously rated on anon [18:13:15] and my previous ratings aren't showing up [18:13:52] [[Geography of Comoros]] should be showing the widget [18:13:53] well, it'll likely take a while to see any load from AF [18:13:54] But it's not [18:16:00] RoanKattouw, Ryan_Lane: I thought we had decided to enable no more than 10% for clicktracking [18:16:13] D'OH [18:16:16] yes [18:16:27] we also decided on how we were going to ramp things up ;) [18:16:27] LotteryOdds is a percentage [18:16:31] I set it to 0.01 [18:16:35] oops [18:16:37] So we're now on 0.01% nowt 1% [18:16:45] RoanKattouw: ok cool [18:16:45] Let's actually go to 0.5% now [18:16:52] ClickTracking is at 0% temporarily at Ryan_Lane 's request [18:17:10] that was the plan ;) [18:17:15] did no one write this stuff down? :) [18:17:23] <^demon> Pfft, see bug 1. [18:17:29] heh [18:17:31] RoanKattouw: i thought we would ramp up to no more than 10% [18:17:53] ClickTracking is going to 2% [18:17:59] *going to go to [18:18:15] Now raising AFT to 0.5% from 0.01% [18:19:21] (And AFT is gonna end up at 2.7%) [18:19:36] 2.7%? [18:19:45] is that 1M? [18:19:46] 2.7% of all ns0 articles [18:19:49] 100k [18:19:55] ahh. ns0 articles :) [18:19:58] Technically it's ~115k [18:20:04] sounds good [18:20:11] (2.7% = 110k , plus 4k categorized pages) [18:21:25] I bumped to 0.5% 4 minutes ago, it should kick in some time in the next 6 mins [18:22:01] ok [18:22:42] Aha [18:22:44] AND [18:22:45] There's a JS error in AFT [18:22:47] Breaking it completely [18:23:12] ouch [18:23:27] Or maybe it's debug-mode-specific [18:24:31] i'm not seeing anything [18:24:57] Aha, here we go [18:24:59] The error is gone [18:25:04] And the tool shows u[ [18:25:07] http://en.wikipedia.org/wiki/Geography_of_Comoros [18:25:40] looks good here [18:26:06] i am not seeing any js errors in ff 4 [18:26:17] Yeah that was just debug mode [18:26:25] You're seeing the tool on that article though, right? [18:26:31] yes [18:27:02] yes i can [18:27:03] Alright [18:27:23] So, couple minutes on 0.5% and not seeing anything on the load graphs [18:27:53] hmm [18:28:20] never mind -- tool wasn't showing up on geography of comoros and now it is [18:30:16] RoanKattouw: load on squids went up some [18:30:25] for 1-min load [18:30:41] That's kinda expected [18:30:44] network went up some as well [18:30:45] API squids? [18:30:45] yeah [18:31:04] Yeah I see it [18:31:11] yeah, you should be watching the squids and the realservers ;) [18:31:33] from an actual load perspective, not much has changed though [18:31:51] which is good. that's what we were expecting :) [18:31:55] No, the traffic bump you saw was more of a fluke [18:32:10] Besides, you'll be happy to hear we're aggressively caching API results in Squid this time [18:32:22] And only doing API calls when the widget is in view [18:32:49] yeah, I asked for those changes ;) [18:32:57] Oh, wasn't aware [18:33:03] I just saw Trevor implementing them [18:33:07] *Ryan_Lane nods [18:33:29] RoanKattouw: can you expand on this "only doing API calls when the widget is in view", I'd like to know the implications on the data we are collecting [18:33:33] Anyway, no visible change, ready to go to 1% (from 0.5%) ? [18:33:37] DarTar: Implications are zero [18:33:49] sounds good to me [18:33:58] RoanKattouw: i'm actually not observing that behavior [18:34:03] DarTar: Basically if the widget is off-page on a long article, we're not loading ratings data into it until you actually scroll down and bring it into view [18:34:31] is this behavior only for users that have not yet rated? [18:34:32] ah, that's interesting. I hadn't asked for that change, but it's a good one :) [18:34:33] howief: I didn't write it and didn't test it, so ... :) [18:34:48] ok [18:34:55] Alright, so going to 1% nwo [18:35:21] RoanKattouw: that's interesting, one of the things we looked at was actually the volume of ratings depending on article length, but from our point of view it shouldn [18:35:21] t [18:35:27] make much of a difference [18:35:38] It really makes no difference at all [18:35:56] We just delay loading of the widget as long as it's off-screen [18:36:06] *Ryan_Lane is a fan of this [18:36:15] sounds great [18:36:21] there should only be a difference if there is a bug and the widget doesn't show [18:37:14] and even then, its not going to hide the widget, just not load the rating data initially [18:37:38] its a subtle thing [18:38:29] and. we could, in userrating-0 cases, load when they click on view page ratings [18:38:32] TrevorParscal: so on a previously rated article, the widget loads and then previous ratings load? [18:38:43] yes [18:38:49] cool [18:39:26] TrevorParscal: quick question-- how are the averages calculated on the widget? [18:39:36] TrevorParscal, RoanKattouw: labels for the ratings in the dashboard do not appear correct - that is, it looks like they're pointing to non-existent messages [18:39:43] Oh, bah [18:39:46] Let me fix that [18:39:49] i know they're moving averages, but I can't remember how many are in the moving average [18:40:09] howief, very carefully [18:40:11] and we'll need to run that maintenance script at some point too [18:40:28] haha [18:40:46] howief, they are the average of all non expired ratings and the count of all ratings ever [18:40:55] awjr: Fixed in the DB [18:41:06] nice thanks RoanKattouw [18:41:08] got it -- and what's the definition of expired? [18:41:25] 30 revisions old [18:41:34] RoanKattouw: how/when do we run that maint script to populate the dashboard table, and how do we cron it? [18:41:41] great thanks! [18:41:49] Ugh, there's that script too [18:42:05] I'll run it later toda [18:42:09] y [18:42:11] Then I'll put it in my crontab [18:42:20] We can worry about setting it up properly after Berlin [18:43:18] RoanKattouw: in your crontab? :) [18:43:18] cool [18:43:20] *Ryan_Lane stabs [18:43:25] except for that :p [18:43:41] RoanKattouw: please put an RT in before you do this, so that we can track this [18:43:49] OK [18:43:59] I know it's not supposed to be in *my* crontab [18:44:12] if there's one thing I dislike, it's site features depending on non-system level user accounts [18:44:15] But with everyone going to Berlin real soon now I'd rather take the easy cop-out [18:44:21] For the next two weeks or so [18:44:27] yeah. that's fine. we can fix it there [18:44:32] but we need to track that there's a hack in place [18:44:36] Will do [18:44:55] ty [18:45:36] Woot the load-data-when-widget-becomes-visible behavior works for me [18:45:39] As does the 1% thing [18:45:44] See http://en.wikipedia.org/wiki/April_12 [18:45:46] :D [18:46:30] So real movement visible on Ganglia [18:48:41] and dropping [18:48:44] eh? you notice a change in ganglia? [18:49:10] spike in cpu on the api squids but it's going back down [18:49:38] I'm not seeing this [18:49:43] you mean that tiny blip? [18:50:17] i suppose that depends on your pov [18:50:25] heh [18:50:28] API Squids network is climbing slightly [18:50:31] Could be a blip [18:50:49] we are also moving into peak [18:51:00] Yeah [18:51:23] Nope, what I saw was probably a blip, it's going back down [18:51:49] I see no noticable change [18:51:58] RoanKattouw will it be showtime any moment now ? [18:51:58] though it could be a day or so till we really see anything [18:52:04] since people have to notice it [18:52:08] and start using it [18:52:29] GerardM-: it's showtime right now [18:52:55] GerardM-: We're still doing AFT, so hang on [18:53:09] We're ramping of the # of articles that get AFT [18:53:14] When that's all done we move to UW [18:53:14] oh, were you not referring to AFT? :) [18:53:20] what's UW? [18:53:25] UploadWizard [18:53:29] Routine bugfix deployment [18:53:31] Ryan_Lane: There's a message from Erik on the Village Pump which might have a follow-up announcement added to it. [18:53:32] ahhh. right [18:53:34] Was supposed to start at 11 [18:53:55] Aaanyway [18:54:03] I'm not seeing any change at all on Ganglia [18:54:12] We've been on 1% for 10-15 mins now [18:54:46] So, what next? 2% , then if that goes well 2.7% (the target)? [18:54:54] sounds reasonable [18:55:37] :) [18:55:56] sounds good [18:55:58] let's do it [18:57:10] There, give it 5-10 mins to kick in [18:57:14] *RoanKattouw looks up an affected page [18:58:03] http://en.wikipedia.org/wiki/Computational_chemistry should get the tool in a few minutse [18:58:08] RoanKattouw: you make aft sounds like a disease [18:58:14] haha [18:58:16] It's viral! [18:58:19] :P [18:58:26] lol [18:58:50] I see a spike in API traffic in now [18:59:02] On the Apaches but not on the Squids [18:59:29] i see the tool on the page [18:59:54] I don't, yet [19:00:23] Maybe you're using secure? [19:00:43] nope, i clicked the link you pasted :p [19:01:17] *TrevorParscal is queueing in airport security [19:01:31] Traffic bump was a fluke [19:01:46] TrevorParscal, how exciting! [19:01:58] Ah I get it now [19:02:31] TrevorParscal: are you gonna opt for the scan or the grope? [19:02:43] I didn't get the scan last time at SFO [19:02:55] I did get it at AMS when I went to England, not sure about my last US visit [19:03:02] RoanKattouw: I can't see AFT on http://en.wikipedia.org/wiki/Computational_chemistry yet [19:03:12] awjr could before I could [19:03:14] I can now [19:03:22] i can see it [19:03:23] alolita: Try refreshing [19:03:35] (or even hard refreshing) [19:03:52] RoanKattouw: can see it now :-) [19:03:57] i had the scan last time i flew out of sfo [19:04:01] API cluster is still going on its merry way [19:04:31] awjr, grope... of course [19:04:48] yeah, everyone wins, i guess [19:05:29] "Oooh, you really do care about security don't you!" [19:05:44] hahaha [19:06:15] Not as much as I care about minimizing the radiation dose my body is subjected to [19:06:21] TrevorParscal, you'd better put your phone away before security see you on hacker irc [19:13:21] Alright, 17 mins in and I'm seeing nothing on Ganglia [19:13:36] Ryan_Lane: Am I good to go to the full 2.7% (now at 2%)? [19:13:43] yep [19:15:20] Done [19:15:33] w00t [19:15:36] howief: We will be at the full 2.7% (= roughly 115k articles) in a few minutes now [19:15:44] awesome! [19:16:05] If that holds up I'll turn click tracking back on at 2% [19:16:26] sounds good [19:17:43] great [19:20:32] RoanKattouw: do you have a page we can check for this latest bump? [19:20:39] Lemme get one [19:21:05] http://en.wikipedia.org/wiki/Economy_of_the_Democratic_Republic_of_the_Congo [19:21:47] showing up for me [19:22:19] Bah, there's a bug in ResourceLoader, the startup module has a Last-Modified that's wrong [19:22:47] In that it doesn't take config changes into accuont [19:23:33] So you may need a refresh or hard refresh to get AFT on those pages [19:23:46] i got it with a hard refresh [19:24:02] Me to [19:24:04] o [19:24:09] Firefox was still 304ing on a soft refrseh [19:24:13] *RoanKattouw files a bug about this [19:25:47] OK I'll work around it by touching a random file [19:26:00] Or, not so random, startup.js [19:28:58] RoanKattouw: still holding up at 2.7% (= roughly 115k articles)? [19:29:12] Yeah looking good [19:29:36] So let's turn click tracking back on, at 2% [19:30:01] That will give us roughly the same volume (actually a tad more) that we got before [19:30:11] 25x lower tracking percentage on ~25x as many pages [19:35:13] neilk_: Is trunk of UploadWizard deployable right now? [19:35:17] (and reviewed) [19:35:34] RoanKattouw: no and no [19:35:44] Brion reviewed it, found some bugs I'm trying to fix [19:35:45] OK... should we deploy at all today? [19:35:53] however, he didn't mark the other changes "ok" either [19:36:02] It's 9:45pm and I have to pick up Trevor from the airport at 9:15am [19:36:10] RoanKattouw: hard to say. There is a more minimal set of changes we could deploy [19:36:22] RoanKattouw: but they aren't marked ok either yet [19:36:27] Hm [19:36:34] sorry [19:36:37] No worries [19:36:41] It's convenient for me in fact [19:36:57] I'll get some time to shower and pack before heading off for an early night [19:37:23] I should leave early tomorrow because traffic will probably be bad [19:37:30] I would like to get the more minimal set of changes deployed [19:37:37] but if you are vetoing it, I understand [19:37:47] I haven't been on the Amsterdam beltway at 8:30 in the morning before but it can't be too good [19:38:19] neilk_: Given the way my eyelids feel right now, I'm saying no. I guess we can get together in Berlin and do it, if ops is cool with that (at least they'll be around :) ) [19:38:35] RoanKattouw: well, ops is one thing, but we sent out announcements too [19:38:41] RoanKattouw: not that this is your problem [19:38:45] heh [19:38:58] and it's hackaton, not deployaton [19:39:17] If it's a quick and uncomplicated set of changes I can do it [19:39:18] let me ask Brion if he can mark some changes okay, they are the simpler changes [19:39:46] Why don't I take a shower and we regroup in 30 minutse [19:41:06] i'm poking some more [19:41:21] howief, alolita, Ryan_Lane, TrevorParscal, Krinkle: AFT is looking good to me. We're ramped up to where we want to be: 2.7% of all main namespace pages plus categorized pages (~115k pages total), click tracking at 2%. Servers don't seem to be breaking a sweat I'm stepping out to take a shower, back in 30 minutes [19:41:28] brion: the error recovery idea isn't working and I'm not sure why [19:41:31] great! [19:41:35] :( [19:41:38] thanks to everyone! [19:41:41] RoanKattouw: AWESOME [19:41:43] brion: however the hacky regex check is relatively uncomplicated [19:41:54] this is super exciting :) [19:42:12] RoanKattouw: Nice work! [19:42:13] When do we ramp up to more articles again? Late May, right? [19:42:13] howief: did you do an announcement on blogs.wikimedia.org [19:42:17] yeah the regexes seem to be running at least, as it does that initial catch [19:42:27] RoanKattouw: yes, when everyone is back from the hackathon [19:42:31] but after the one the regexes don't catch, maybe the state's getting messed up somehow [19:42:33] RoanKattouw: yes late May [19:42:35] alolita: not yet [19:42:46] howief, DarTar: Also, I can get you guys a list of all pages it applies to if you want it. Send me an e-mail and you'll get it tomorrow [19:43:01] nimish_g: Your request for the dump thingy will also have to wait till tomorrow [19:43:02] brion: the real problem is that I have two error libraries contending... I tried to extend one and ran out of time, thought a hack would work, it's not working. [19:43:05] i'm thinking we just use erik's wiki-en msg with some modifications [19:43:11] RoanKattouw: awesome! [19:43:26] howief: that's a good plan [19:43:27] fantastic [19:43:31] brion: but the regex checks are pretty vanilla and would REALLY help [19:43:42] neilk_, is it a regression from the previous state of things? [19:43:53] RoanKattouw: okie [19:43:54] or did it previously die and not recover ever, and now it's not really any worse [19:44:11] brion: well, it's not worse [19:44:16] \o/ [19:44:32] that's progress of a sort ;) [19:44:55] we'll see if we can figure it out then but i wouldn't let it stop your deploys then [19:45:01] howief: what's the plan for a blog post announcing AFT? Shall we start a draft somewhere? [19:45:14] ok, Roan said he wants to take a shower & come back for UW deploy [19:45:17] DarTar: there is a draft that Erik had put together [19:45:26] DarTar: we should use that [19:45:26] RoanKattouw_away: lmk when you'll be running that maintenance script for AFT so i can be sure to be around [19:45:33] brion: my problem is that I need someone to actually mark the changes as ok [19:45:49] brion: if it's not you I can get s.o. else [19:45:54] neilk_, well i'm marking away [19:45:55] DarTar: alolita i just sent guillom a message to see if he has time to help with the blog post [19:45:56] see in #mediawiki [19:46:01] but yes, i do think we should keep things very simple [19:46:03] kaldari: http://www.mediawiki.org/wiki/Special:Code/MediaWiki/87761 [19:46:09] ok, nm [19:46:12] ok cool [19:46:13] and use erik's post [19:47:00] it's probably late for guillom tonight but we can target a blog post tomorrow [19:47:20] oh shoot - i keep on forgetting about the time differences [19:48:55] it's just before 10pm for him [19:51:06] ok -- let's try to get something out later today then [19:51:16] we'll bother guillom only if we need to [19:52:42] howief: it will be good if guillom is available [19:53:00] yes it would be, but it's probably late for him [19:57:05] I can blog as well [19:57:15] am waiting actually for it [19:57:29] it is as late for me as for guillom [19:57:48] I dunno, maybe this is deployable [19:58:10] brion: that error seems to occur only if someone is really determined to ignore both sets of error messages [19:58:17] brion: 3x [19:58:29] GerardM: Hi! [19:59:01] GerardM: Can you coordinate with Howie and see if you can help :-) [19:59:15] Hoi Alolita ... quite happy to do this [19:59:30] I have an email to foundation commons and Wikipedia-l ready in concept [19:59:56] GerardM: great - thanks! you can get deeper details from howief [20:04:25] howief, DarTar , it's definitely too late for me, it's 10:12pm and I'm going to bed. [20:21:27] awjr: Alright, let's do your maintenance script [20:21:38] sure now's good for me [20:22:04] Running on testwiki first [20:22:09] k [20:22:12] That was fast [20:22:17] Can you check the dashboard on test? [20:22:42] dashboard is disabled on test [20:22:56] lemme enable it [20:24:05] OK [20:24:36] RoanKattouw: can you fix the labels on testwiki as well? [20:24:51] Right, sorry [20:25:00] no worries - what causes that? [20:25:31] Done [20:25:36] Out-of-date DB contents [20:25:41] INSERT INTO /*_*/article_feedback_ratings (aar_rating) VALUES [20:25:42] ('articlefeedback-field-trustworthy-label'), ('articlefeedback-field-objective-label'), [20:25:44] ('articlefeedback-field-complete-label'), ('articlefeedback-field-wellwritten-label'); [20:25:46] Used to be something else [20:25:46] ah [20:26:20] ok - it looks good, but we should do some more rating (we need > 10 ratings in last 24 hours on a few articles) [20:27:28] hey Krinkle, brion, you marked http://www.mediawiki.org/wiki/Special:Code/MediaWiki/87722 as a fixme, but what is the preferred fix? [20:27:41] RoanKattouw: i just added some more ratings for the main page - can you run the maint script again? [20:27:42] Krinkle, brion - it's a hack to comment out behaviour that kills it in IE [20:27:45] that should give us data in both tables [20:28:08] Writing data to article_feedback_stats_highs_lows ... [20:28:09] Inserted 2 rows [20:28:14] Krinkle, brion - if there's a super easy way to switch to regular mw.log, I guess I could do that [20:28:18] neilk_: I didn't mark it fixme. I suggesting removing it at some point. [20:28:26] neilk_, i'd say let's switch it to a todo and mark ok for now [20:28:35] todo == switching it to the common mw.log code when time permits [20:28:40] uhoh [20:28:48] highs/lows appear... reversed [20:28:59] http://test.wikipedia.org/wiki/Special:ArticleFeedback [20:29:48] Hm... clicktrackingsession=MphSiYHkNxoBtpbh1KaWZsMhh243nLeo4&clicktrackingevent=ext.vector.sectionEditLinks-bucket%3A1%402-tab-save in the URL ? [20:29:57] Yes [20:30:09] You clicked an edit/login/createaccount link from an AFT call-to-action? [20:30:23] hehehe codejinx [20:30:26] #Comment by Brion VIBBER (Talk | contribs | block) 13:37, 9 May 2011 [reply] [20:30:30] #Comment by NeilK (Talk | contribs | block) 13:37, 9 May 2011 [reply] [20:30:32] No, I clicked edit on [[Sandbox]] @ testwiki [20:30:33] awjr: Just change the messages around? :D [20:30:39] hahahah [20:30:41] (Or find the real bug) [20:30:46] i'm opting for the latter [20:31:01] weird i havent seen this locally [20:32:14] Can someone re-run the script ? Right now there's only 1 row in each table, hard to see if the order is working properly. [20:32:16] I added a third page [20:32:39] Rerunning [20:33:26] oh i think i see the problem [20:39:44] RoanKattouw: i think it was not selecting ratings in the appropriate order before caching from the maint script - should be fixed in r87768 [20:39:45] http://www.mediawiki.org/wiki/Special:Code/MediaWiki/87768 [20:40:17] brion: thanks for the reviews & comments -- the more optional ones I'll leave off till after deploy [20:41:06] brion: or, if you're waiting on that fix for 87764 I'll just do it now [20:41:25] once that's on test, would you mind truncating article_feedback_stats_highs_lows and r unning the script again? [20:41:29] yeah i'm todo'izing most of em [20:42:06] that and 87410 are both small things that are either fine for now or might need a small tweak [20:42:27] depending on whether the error condition is actually possible :D [20:44:01] awjr: Done except the truncation part [20:45:13] looks correct now [20:45:54] RoanKattouw: do you think we can still deploy tonight? do I have time to step out for a few minutes? [20:46:41] brion: crap you are right about the exception. very good catch [20:46:46] neilk_: we should try to deploy today if possible [20:46:49] \o/ [20:47:02] neilk_: We can deploy when ever you want [20:47:05] brion: *I* wrote that and I forgot [20:47:06] But sooner==better [20:47:16] :D [20:47:16] anytime you want [20:47:17] awjr: OK so ready to run on enwiki? [20:47:21] yep [20:47:34] Alright running [20:47:38] And merging stuff for Neil [20:47:38] Woot [20:47:40] neilk_: Trunk deploy? [20:48:03] *awjr crosses fingers and hides [20:48:09] awjr: Done [20:48:15] That was quick and painless [20:48:19] The slowest part was inserting the 100 rows [20:48:26] Because writes on s1 are godawfully slow [20:48:27] I have a concept for a blogpost [20:48:29] yay looks like it works [20:48:41] http://en.wikipedia.org/wiki/Special:ArticleFeedback [20:48:57] howief: http://en.wikipedia.org/wiki/Special:ArticleFeedback [20:49:05] RoanKattouw: yeah, hold off like 2 minutes [20:49:06] for UploadWizard [20:49:22] neilk_: OK just tell me when [20:49:59] There's no 'talk page' for special pages, right? [20:50:42] no [20:50:59] This copy on the AFT dashboard is rather misleading: 'Please provide feedback on the talk page.' [20:51:04] since there's no talk page... [20:51:08] howief: ^ [20:51:43] wow [[The Pirate Bay]] has gotten no love in the last 24 hours [20:56:48] Just making sure but ahm.. blocked users can't rate, right ? [20:59:00] RoanKattouw: thanks for helping get that out :) [20:59:07] Sure [20:59:10] neilk_: ETA? [20:59:17] now :) [20:59:29] you can trunk-deploy [21:00:03] Alright [21:01:41] RoanKattouw: did you make a cronjob for that maint script? [21:01:55] Not yet [21:02:00] k [21:02:02] How frequently would you like it to run? [21:02:06] daily [21:02:13] It runs very quickly, it looks like hourly would be feasible [21:02:25] s/very/fairly [21:02:37] because of how it's written, it needs to be daily [21:02:44] neilk_: Trunk UW is on test [21:02:59] awjr: OK. What time? I'm assuming it needs to be later than the last time it ran today? [21:03:12] actually i made that up [21:03:18] lol [21:03:34] Can I go ahead and cron it for 15 past the hour then? [21:03:45] (as in every hour0 [21:04:05] sure [21:04:07] sounds good [21:04:47] we should keep an eye on its performance as the #s of ratings increase [21:04:49] Alright, it's in my crontab [21:04:51] :D [21:04:52] So it should run in 3 mins [21:09:42] RoanKattouw: awesome stuff [21:11:09] if a talk page can't be set up for special pages we can just point to the main Talk page of the AFT on mw.org [21:11:59] DarTar: we'll point users to http://www.mediawiki.org/wiki/Talk:Article_feedback [21:12:01] awjr: Hmm looks like the cron job failed [21:12:03] RoanKattouw: UW is deployable I think [21:12:04] :( [21:12:21] neilk_: OK. It's on test already, do you want to go to production now? [21:12:41] RoanKattouw: yes that's what I meant. [21:12:43] scappable? [21:12:44] RoanKattouw: when you have a sec - any idea why? [21:12:51] awjr: Fail, typo in crontabl, fixing [21:12:55] neilk_: Will deploy [21:13:21] Pro tip: put php foo.php --wiki=enwiki in your crontab, not just foo.php --wiki=enwiki [21:13:44] neilk_: Scapping [21:13:57] awjr: Cron job set for 25 past the hour [21:14:04] great [21:16:47] RoanKattouw, awjr - is it also as expected that the default sort on the dashboard is by average *ASC* ? [21:17:27] good q for Trevor or Krinkle, but if i remember our convo from friday correctly, that is correct [21:17:31] DarTar: ^ [21:17:52] although it probably makes more sense to do desc for highs and asc for lows [21:17:55] ok I may have missed that [21:17:59] agreed [21:18:48] Indeed, desc for high, asc for low imho [21:19:36] RoanKattouw: did that cronjob run alright? [21:19:51] Looks like [21:19:58] select distinct afshl_ts from article_feedback_stats_highs_lows; [21:20:03] | 20110509205549 | [21:20:05] | 20110509212025 | [21:20:06] | 20110509212502 | [21:20:12] Maybe we should periodically clear out the old stuff? [21:20:37] yeesh if we're doing this every hour... [21:20:46] that's 100 new rows every hour [21:21:18] we had chatted a bit about the value of hanging on to the data and periodically pruning but never got into specifics [21:21:32] yeah [21:21:55] at some point it may be nice to have something like a permalink [21:22:08] but we should think through those use cases [21:22:16] I expect the 24h fluctuations will be too high to have the same articles for two consecutive days on the list [21:22:21] Permalinks would be nice [21:22:27] And trivial to implemet [21:22:33] hmm. . . [21:22:36] But since the data is persistent maybe hourly isn't a very good idea [21:22:41] i'm still seeing some weirdness in the data [21:22:44] (Right now it's configured to refresh hourly) [21:22:59] RoanKattouw: yes, that's one of the nice to have things we mentioned at the end of the AFT doc [21:23:03] howief: i agree, some of these are suspect [21:23:09] yeah [21:23:16] kaka is still showing 5.0's all across [21:23:40] something similar to ohloh's factoids (permalinked, timestamped stats like, "this article was top rated article on this date") [21:23:51] Maybe awjr's calculations don't take expiry into account or something [21:24:18] expiry? [21:24:20] of? [21:24:42] i don't think expiry would be the cause of the issues [21:25:06] if anything, taking expiry into account would make a 5.0 more likely [21:25:09] of ratings [21:25:21] wait, ratings expire? [21:25:32] Oh, here goes [21:25:36] ? [21:25:38] It only accounts for the past 24 hrs [21:25:41] yes [21:25:48] It avgs the ratings of the past 24h [21:25:48] but all ratings within past 24 hours [21:25:55] Of course that won't necessarily match what the AFT widget says [21:25:57] a 5.0 average is extremely unlikely [21:26:05] unless someone's gaming the article [21:26:07] based on the data we've seen [21:26:08] Want me to manually inspect the data? [21:26:12] would you mind? [21:26:18] Which page? [21:26:20] RoanKattouw: is that something i'm able to do as well? [21:26:27] Of course [21:26:28] http://en.wikipedia.org/wiki/Kak%C3%A1 [21:26:39] shoot [21:26:43] awjr: I'll show you later when I'm not tired. You'll be in Berlin,right? [21:26:47] yes [21:26:49] RoanKattouw: thanks, looks like it's working. [21:26:50] and that would be awesome :D [21:28:25] We do indeed have all fives [21:28:35] really? [21:28:38] 4 ratings in the past 24h on that page [21:28:42] All were 5,5,5,5 [21:28:42] hmmm [21:28:43] on 4 ratings? [21:28:47] er, only 4 ratings? [21:28:52] strange [21:28:54] there should be at least 10 ratings in the last 24 hours [21:29:03] in order for that article to show up [21:29:15] also, the article page is showing 80+ ratings [21:30:21] http://dpaste.org/95nw/ [21:30:34] (For the benefit of both awjr for the queries and howief for the results) [21:30:40] that is indeed all 5's [21:30:42] haha [21:30:47] I'm afraid awjr is miscounting [21:30:51] 4 ratings = 16 rwos [21:31:02] hmmm [21:31:09] CABAL of the "Article feedback"-project!!! I call thee out and challange thee! [21:31:12] there should be more than 4 total ratings, no? [21:31:24] RoanKattouw: according to the spec, one rating is... one rating, not 4 [21:31:31] Yes [21:31:33] In the table, it's four [21:31:33] so one rating = 1 row [21:31:40] *Bensin looks around... [21:31:44] Who steps up? [21:31:48] They're talking about it, Bensin. [21:31:59] It seems we have different notions of what "one rating" means [21:32:06] indeed [21:32:13] But most likely what you want to say would be best said on Village pump:Tech. [21:32:20] but according to the spec document, one rating would = one row [21:32:34] cabal, eh? [21:32:46] [21:32:48] Amgine: Thanks. If the talk is here, I'll stay here. [21:32:51] I haven't seen that spec, but I can tell you that that would be slightly counter-intuitive [21:33:06] i'm actually a little more worried about the total number of ratings shown on the article page at this point [21:33:07] how / when is the upload button on Commons changed to the uploadwizard ? [21:33:29] awjr: Well it has all the markings of a shadowy project... [21:33:33] howief: So this article has been rated 4 times, and each time the rater filled out all four entries. In the DB that's 4*4=16 rows, so with Arthur's script that's >10 and the page qualifies [21:33:39] (to be honest, 80 ratings within a few hours seems high for kaka, unless either it's part of the previous batch or he's much more popular than i'd assumed) [21:33:50] No, that's total unexpired ratings [21:34:03] 80+ ratings over the lifetime of the article [21:34:06] Bensin: its been running on en.WN for a year or so, iirc. [21:34:11] or rather, over the lifetime of the tool being available on the article [21:34:15] I see 111 ratings in its lifetime [21:34:22] oh ok [21:34:23] got it [21:34:31] so this one was actually part of the previous batch [21:34:39] 20110324011639 [21:34:42] Is the oldest rating [21:34:49] so in order to qualify for the dashboard [21:34:52] it should actulaly be 10 rows [21:35:11] Amgine, btw readerFeedback and enwikipedia's article feedback tool are separate extensions (enwikipedia's is nicer ;) [21:35:11] 10 rows not 10 distinct ratings? [21:35:11] from the spec doc: "List will be calculated based on all articles that have received over 10 ratings in the previous day (this is actually rating records, i.e., if a user rates only 1 of the 4 categories, this still counts as a rating)" [21:35:22] OK [21:35:29] Amgine: Surprising it's so underdeveloped and shapeless then.. [21:35:36] But if the user rates 3 of the 4 categories, does that count as 3 ratings or still as 1? [21:35:45] Bawolff: I know... but the project has been developing even longer... strat wiki. [21:36:00] RoanKattouw: that counts as one rating [21:36:09] Right [21:36:17] oh [21:36:19] awjr: So, populateAFStatistics.php line 92 [21:36:25] $rating_count += count( $rating ); [21:36:26] then i completely mis-read that [21:36:30] I guess that should be $rating_count++; [21:36:32] or, misinterpreted that [21:36:59] >_< [21:36:59] brb [21:37:19] incidentally Kak?? is definitely not one of our most viewed articles: http://stats.grok.se/en/201105/Kak?? [21:37:57] I have a request that the rating-box has a link to the projects page so that editors wondering what that weird box is can check it out... Can someone fix that? [21:38:01] DarTar: makes sense [21:38:04] oh, that's clearly a fan pumping the stats. [21:38:09] [21:38:32] three browsers anonymously + user account [21:39:01] yeah if this dashboard proves useful to the community, we'll have to figure out how to deal with ppl juking the stats [21:39:29] My request was put forward by myself and two other editors here: https://secure.wikimedia.org/wikipedia/mediawiki/wiki/Talk:Article_feedback/Public_Policy_Pilot/Workgroup#A_link_to_project_page [21:39:30] gaming happening within hours of the deployment, yay! [21:39:42] I think it's healthy [21:39:52] :) [21:39:52] Actually it looks like this may have happened earlier today [21:39:53] ok, so this new revelation about what a 'rating' is makes a lot more sense :p [21:39:58] shows ppl are paying attention at least [21:40:01] Kaka already had the widget [21:40:15] attention as to how to game it. [21:40:18] maybe this user found out about the dashboard? [21:40:25] but at least it's hard to "mass game" [21:40:42] howief: No. timestamp of rating: 20110509145553 [21:40:43] RoanKattouw: the safe thing to do then is consider a page_id/anon_token combo a 'rating'? [21:40:55] awjr: Look at the index on the table [21:41:03] there are many ways to build a robust algo, but we need to start a separate discussion for that [21:41:04] It's something like revid and user [21:41:08] interesting [21:41:14] (the unique index, I mean) [21:43:31] ah yes [21:44:01] this is amusing, i thought i had properly clarified this last week when i actually wrote this :p [21:45:32] :) [21:47:15] jorm: I have a request that the rating-box has a link to the projects page so that editors wondering what it is can check it out... Can you help me with this? [21:48:22] http://ultimategerardm.blogspot.com/2011/05/uploadwizard-became-default-upload-tool.html [21:48:44] Bensin, I wish i could help you but I am not on the project as such. [21:49:17] Bensin: were you thinking a link to the mediawiki page? [21:49:44] (http://www.mediawiki.org/wiki/Article_feedback) [21:50:13] howief: Yes. Or wherever one is expected to catch up on what is going on... [21:50:30] yeah that may be a good idea [21:50:39] at least for the pilot phase [21:50:49] As of now it is just a box that appears on article without being represented in the wiki code. [21:51:01] Thank you! [21:51:22] Maybe this project is more than a praying mantis that keeps mating even after you bite its head off :-) [21:51:34] i'll check with Trevor to see if we can work it in [21:52:12] though I"m not sure we'd want a link to the mediawiki page if/when this is deployed across all of engligh wikipedia [21:52:14] howief: i might not be able to get this fixed/deployed by eod today - should i disable the dashboard in the interim? [21:52:26] the feature is targeted more at readers than at editors [21:52:46] awjr: when do you think you'd be able to pick this up? [21:52:55] Well, it's the editors that will notice it first... [21:53:02] Bensin: yes [21:53:07] assuming i get this code reviewed in time i should be able to have it out before i leave for berlin [21:53:17] Bensin: i think this is a good idea for now [21:53:40] but for full deployment, i don't think the mediawiki page is the right one since many readers will click on the link as well [21:53:41] or we can just leave it on and roll out the fix asap [21:53:54] let's disable for now [21:53:57] k [21:56:27] disabled [21:57:24] howief: According to Erik Moeller, one of the intentions of the tool is to be "an entry vector for other forms of engagement". A link to read more about the ratings-box itself is a way to engage the reader in the ratings project. Name the link "Read more about the ratings project" or something and let curious people click it all they want. [21:57:25] thanks awjr [21:57:35] no sweat [21:58:07] Bensin: yes, but as written, it will likely confuse readers more than help [21:58:46] all i'm saying is that we know that readers will click on the link if it's there [21:58:59] so the page we send them to needs to speak to both readers and editors [22:03:17] You are an editor, I presume; you already have a much higher passing understanding of the way it works. [22:03:35] at any rate, i think it's a good idea to include a link in the widget during the 100k test [22:03:53] a general help link upon full rollout is probably a good thing too, as long as it links to a sane page for both readers and editors [22:03:58] jorm: And yet I had a HARD time finding this channel and put forth my opinion :-) [22:04:02] (help or more info link) [22:04:02] I don't disagree, and I think I had one in my original designs, howie. But it's been a while since I worked on this. [22:04:22] jorm: you're prob right [22:05:07] Perhaps we need to add a line to the topic like "By speaking here you are aware of the chances being quotes in BugZilla's quips" [22:05:13] I'm trying to come up with a way to make our designs and projects more open but it's hard. [22:05:16] quoted* [22:05:42] you get weird pushbacks in weird areas. [22:15:57] howief: if i'm reading things right, we are considering a unique 'set' of ratings (or 'a rating' in the parlance of the spec) for a particular revision id + user id (or anon token for anon users) - does that sound correct to you? [22:22:11] awjr: that sounds right [22:22:33] i think you used the word "record" previously [22:22:42] which is the right concept [22:23:51] hahaha i probably meant database record [22:24:02] like, a row in the database [22:24:06] which = 1 rating [22:24:14] and this defines whether or not an article is "eligible", but then the calculation is done on total? [22:24:17] yes [22:24:30] so every time a user hits submit is another way of looking at it [22:24:55] yeah - i'm now thinking of this as a 'rating set' [22:25:11] 1 rating set = the collection of ratings that get stored when a user hits submit [22:25:12] jorm: calculation is done over all ratings in past 24 hours [22:25:21] awjr: yes, that sounds right [22:25:24] great [22:25:47] oh, so sum(rating_id.value) / count(rating_id) [22:25:51] (i.e., you need 10 ratings sets to quality, but the average shown is the average of all ratings in past 24 hours) [22:26:00] over past 24 hours [22:26:03] okay. so that's how i would have envisioned it. [22:26:14] eligibility determined by activity; values determined by value. [22:26:16] haha yeah, that makes a lot more sense. [22:26:19] :) [22:59:28] Bug triage is starting now: http://etherpad.wikimedia.org/BugTriage