[17:04:41] dapatrick: I'm here, whenever you're ready [17:04:56] I'm in the Hangout already. [17:38:18] http://arxiv.org/abs/1507.06893 [17:39:53] "Do Open Source Software Developers Listen to Their Users?" [17:39:57] (tl;dr: "No.") [17:40:48] "Another limitation of this study is its relatively small sample size. Although we sent our survey to a [17:40:51] considerable number of OSS developers in 19 different projects of software, we received only 72 responses." [17:41:05] IIRC I read that paper ages ago [17:41:10] First Monday: Peer-Reviewed Open Journal on the Internet, 17(3):1-9, 2012 [17:41:39] http://firstmonday.org/htbin/cgiwrap/bin/ojs/index.php/fm/article/view/3640/3171 [17:41:46] for the figures [17:41:56] oh, thanks [17:42:13] Should probably be reviewed in https://meta.wikimedia.org/wiki/Research:Newsletter either way [17:42:15] "/htbin/cgiwrap/bin/ojs/index.php/" - for a digital first open access journal..... [17:42:40] "Bødker, et al. [6], however, observe that OSS developers need to have a thorough realization of user expectations." [17:43:35] that's hard to achieve. half of our devs or more don't edit regularly. and yesterday i asked if hovercards isn't basically equivalent to the navpopups gadget, which is apparently a lolwut gaff no editor worth his or her salt would make [17:44:16] since navpopups is apparently NORAD for wikipedia [17:44:33] ori: pretty sure quiddity informed them of navpopups during their development, (which kind of just reaffirms your point) [17:45:07] * greg-g remembers those task/bug discussions [17:45:18] now that i understand it better, hovercards really seems like a mistake [17:45:33] it's a power-user feature for non-power-users [17:45:56] There's some detailed context to that... I could explain faster/more in voice... [17:46:14] v.... voice? [17:46:30] like, Speech-to-Text? [17:46:33] * ori sets mode +v for quiddity [17:55:36] greg-g, ori, But basically, the (~3) volunteer devs who currently maintain Navpopups - which is a giant ball of frequently broken duck-tape and frustration - really want someone else to properly transform it into a robust extension, with full central i18n etc. I've been vaguely aware of this for years. [17:55:37] I had made this 30 second nutshell video, intending to send it to various mailing lists, in an attempt to find interest in extensionizing Navpopups, https://commons.wikimedia.org/wiki/File:Navigation_popups_quick_tour.ogv [17:55:37] but a few *hours* after I made it, and before sending, I learned about the idea of Hovercards (via a comment in #wikimedia-design). So! I talked extensively with prateek, and he understood, and hence he built Hovercard's architecture (as much as possible) with the intent that it could be expanded later, with a "simple" mode for 'readers', and an "advanced" mode for editors, which will (hopefully, someday) be [17:55:40] a near-feature-parity port of Navpopups. With stability/scalability/internationalizability for all. [17:55:47] or something like that. I need more coffee. [17:56:30] neat, I wasn't aware of the idea to have an advanced mode to hovercards [17:59:24] which would only require it being an actually maintained extension that any PM cared about... :/ [17:59:34] community tech! [17:59:39] that's the new mantra right? [17:59:41] dude [17:59:53] (no work has been done on the advanced mode, as yet, but the architecture should be amenable to it) [18:00:04] * quiddity twitches [18:00:04] sorry, too soon? [18:00:20] "what do we need release engineers for anyway?" [18:00:34] don't poke the bear [18:00:45] I love you, bryan [18:00:55] <3 to you too greg-g :) [18:02:42] * quiddity wanders off to find some industrial strength coffee. It's clearly going to be a long friday. [18:39:08] bd808 and greg-g you guys need your own sitcom [18:39:44] "the hippie and the punk"? Starts next week on FX [18:40:12] http://www.digitaljournal.com/img/4/2/2/7/5/6/i/1/1/1/p-medium/laverne-and-shirley.jpg [18:40:47] Laverne 4eva [18:41:03] Shirley was a square [18:41:17] L7 square man [18:41:40] heh [18:45:09] "they come from different worlds and some see them as arch enemies, but deep down, they share more than meets the eye" is that good enough for the promo? [20:10:33] ori: are you going to send out the announcement about sajax? or should I write something up? [20:11:39] legoktm: damn it, I forgot. Are you offering to in good faith, or just registering your annoyance? If the former, yeah, would be awesome, and I'd owe you one. If the latter, I'll do it right now. [20:12:12] uh, good faith :) [20:12:33] <3 [20:17:54] bd808: any objections against extending https://phabricator.wikimedia.org/T91701 to cover response times? [20:18:31] I figure if login did not stop working but started taking ten times longer, that's something that we at the very least would want to know about [20:19:31] hmmm... what would you measure the time of? The full response somehow? [20:20:46] I kind of agree that we'd want to know but I'm not sure that it is easily computable in a non-jittery way [20:21:56] and we already made login take 10x longer on purpose when we added good hashing [20:23:26] Do we send apache access logs into hadoop? If we did we could really get that sort of metric from there [20:23:40] I'm not sure we do though [20:26:08] ori: http://fpaste.org/252839/38979144/raw/ how does that look? [20:26:45] legoktm: great, thanks very much [20:27:12] predicted response: OMG YOU KILLED EVERYTHING!!! REVERT NOW! (from 6 people total) [20:30:05] yes, those logs go to hadoop, we could easily paint them via XAnalytics; I have no idea how much work a Hadoop-based dashboard is though [20:30:08] see also https://phabricator.wikimedia.org/T102079 [20:38:11] bd808: btw any idea what is the weird sawtooth pattern for login on https://grafana.wikimedia.org/#/dashboard/db/authentication-metrics ? [20:38:37] is that an artifact of how statsd works, dumping data every 5s or something like that? [20:58:29] tgr: yeah those are per time period counts, not cumulative totals [20:59:47] I would expect them to smooth out when you move past group0 and get more traffic [21:00:13] testwiki logins are probably mostly jobs from jenkins [21:01:13] Once the new named error buckets show up we should remember to have godog or ori rm the old buckets [22:04:36] hey dapatrick, I need to schedule some security trainings this quarter-- would be great for you to attend to field questions. Are you around 8/21 and 8/24? [22:20:59] ori: Do you have https://gdash.wikimedia.org/dashboards/frontend/ per project? I really wonder how that looks for Wikidata only [22:21:13] especially because I think a recent change of ours made it significantly worse [22:22:10] hoo: no, but navigation timing samples are also recorded in a database as individual records, so it is possible to produce a graph for wikidata [22:22:15] is there a particular metric you are interested in? [22:23:30] mediaWikiLoadStart to document.onload sounds ok [22:23:45] But I'm not really to sure about things yet [22:24:06] what is the time frame? [22:24:29] i need to do something like this fairly frequently, so I have some python scripts in my home dir that I could adapt for this fairly quickly [22:24:51] I deployed the change that probably made stuff worse at 8pm UTC [22:25:02] so a few hours before that till now would suffice, I guess [22:25:16] It's probably so bad that it's easy to notice [22:25:26] ok, on it [22:52:45] hoo: the query is very slow to execute, despite only having to scan 15m rows, and having an index to use, probably because the table is gigantic... still on it. [22:54:21] csteipp: Yep, I'm around those dates. [22:54:54] ok, good to know :) [23:06:47] hoo: we don't have a ton of samples, it looks like, so maybe not enough to see a trend... here's what i got https://dpaste.de/Dgc0/raw [23:08:29] mh... is that second column the sample number? [23:09:50] second number is domcomplete [23:12:58] not enough samples to be useful, it looks like [23:13:39] :/ [23:14:32] hoo: what do you suspect happened? what is the suspicious deployment? [23:16:03] ori: https://gerrit.wikimedia.org/r/#/c/230154/ [23:16:22] The spinner didn't appear before that change (that never really worked AFAIR) [23:16:46] I'm trying to get rid of that spinner [23:16:53] https://phabricator.wikimedia.org/T108365 [23:18:50] it's broken [23:19:04] 'if ( $ ) ' is throwing a reference error [23:19:41] oh, that's cached HTML [23:19:45] which you updated [23:19:45] nm [23:21:24] yeah [23:21:34] but I think having it is bad either [23:22:16] when is it meant to be invoked? [23:22:24] i.e., how can i make it appear? [23:22:42] https://www.wikidata.org/wiki/Special:Random You will see it all the time [23:23:58] * ori doesn't [23:24:28] You need to be logged in to not ge tthe broken JS, I guess [23:24:46] i am [23:24:53] mh, seems it doesn't work in WebKit browsers [23:24:57] try Firefox [23:25:18] It indeed only appears in Firefox for me [23:25:29] I saw it for the first page I hit but not after than [23:25:31] *that [23:25:43] I see the spinner [23:26:17] I do see a ton of dom reflows on each page [23:26:20] bd808: True... I see it once if I go inprivate [23:26:27] Yeah :( [23:27:03] Not sure what exactly happened, but I think wmf17 brought us considerable performance regressions [23:28:49] * ori not seeing it in ff nightly either [23:33:23] well, on the plus side [23:33:40] if the spinner only appears for logged-in users on certain versions of firefox [23:33:46] it's probably not hurting site performance a whole lot :P [23:33:57] Uhm... yay? :D [23:34:19] Maybe I'll try to reduce the content flashes later on [23:45:52] yeah, there are a lot [23:46:30] and they're quite distracting