[00:49:41] <maxbinder>	 dbrant|brb: mdholloway|afk niedzielski-afk bearND|afk I made a sprint board for 82, lemme know if there is any issue. Sorry for the delay.
[06:54:20] <joakino>	 morning
[07:43:33] <joakino>	 setting up mediawiki-vagrant is fun!
[08:17:59] <phuedx>	 morning
[08:18:02] <phuedx>	 joakino: again?
[08:21:39] <joakino>	 phuedx: for my big laptop, i'm going to try working with a bigger screen on my desk, and keep the air for couch/coffeeshop/travels
[08:21:57] <phuedx>	 that's exactly what i ended up doing
[08:25:42] <joakino>	 its all phabricators fault
[08:25:53] <joakino>	 it needs so much space
[08:25:55] <joakino>	 haha
[11:00:10] <phuedx>	 breaking for lunch
[16:40:50] <joakino>	 phuedx: what should we do about https://phabricator.wikimedia.org/T127250#2288422 ?
[16:41:05] <joakino>	 is the problem if the feature is enabled but then disabled?
[16:50:27] <phuedx>	 joakino: yes
[16:50:41] <phuedx>	 joakino: i don't think that it's a big deal
[16:50:51] <phuedx>	 the data is namespaced-ish "wgMF"
[16:51:04] <phuedx>	 but we should just document the decision to let it drop out of the cache naturally
[16:51:17] <phuedx>	 i originally thought that we should clean up after ourselves aggressively
[16:51:22] <phuedx>	 but i don't think that it's a big deal
[16:52:00] <joakino>	 phuedx: i don't think so either in this case. does the parser cache expire by time too? or only with new versions?
[16:52:29] <phuedx>	 i believe it has an expiration time set, yes
[16:53:04] <phuedx>	 i'm in the standerp hangout btw
[16:55:10] <niedzielski>	 bearND mdholloway: o/ in the content service's lead section response, what is the difference between displaytitle and normalizedtitle?
[16:56:24] <bearND>	 niedzielski: we get these properties from the mobileview API. This is not MCS specific
[17:00:02] <bearND>	 niedzielski: https://www.mediawiki.org/wiki/API:Query#Title_normalization
[17:01:29] <niedzielski>	 bearND: does it make sense in the content service? mobile-sections-lead/barack_Obama yields identical normalizedtitle and displaytitle (Barack Obama)
[17:10:46] <bearND>	 niedzielski: I'm thinking we could potentially get rid of normalizedtitle. I think it's only used in PageLead.adjustPageTitle(). I think displaytitle is what an editor could set and this is what the app displays.
[17:16:11] <bearND>	 niedzielski: or maybe not. normalizedtitle is used when a page gets redirected, probably for history entries
[17:16:39] <niedzielski>	 bearND: that would be good. we could just call it "title" and the client would never have to worry about manipulations might have been performed on it. i don't think the service provides a means to get the original title used (e.g., obama) and from what i can tell, normalizedtitle and displaytitle are always identical.
[17:17:25] <bearND>	 niedzielski: is it different from when we use MW API?
[17:18:49] <niedzielski>	 bearND: i don't think so. we get the redirected field but it's also identical to displaytitle
[17:20:14] <bearND>	 niedzielski: yes, i think having those two differ is rare, but it can happen. Sometimes editors like to add some HTML tags to displaytitle, to italicize portions of the title
[17:20:58] <niedzielski>	 bearND: doh, i see that now
[17:22:31] <bearND>	 niedzielski: tl;dr: we still need the normalized title for API requests, and display title for display
[17:24:15] <bearND>	 niedzielski: here are some examples of displaytitle munging: https://en.wikipedia.org/wiki/Template:DISPLAYTITLE
[17:24:59] <niedzielski>	 bearND: nice
[17:26:08] <jdlrobson>	 phuedx: thoughts around doing an A/B test by using a pseudo random number generator to turn on/off lazy loaded images for 50% of page views?
[17:26:11] <jdlrobson>	 does that sound crazy?
[17:31:26] <phuedx>	 jdlrobson: as bblack says, that requires everyone hitting the servers and not just the caches
[17:31:43] <phuedx>	 prngs to bucket users i'm fine with
[17:31:49] <phuedx>	 but i'm not sure where you're wanting to do the bucketing
[17:32:11] <jdlrobson>	 we need 50% of page views to have lazy loaded images enabled and 50% to not have it
[17:32:42] <jdlrobson>	 given MobileFormatter result gets cached in varnish layer I was curious if the use of a random number in php to eanble lazy loaded images would provide this result
[17:42:56] <phuedx>	 jdlrobson: how is the varnish cache split for anons?
[17:44:05] <jdlrobson>	 phuedx: that i'm not sure about.
[17:56:47] <dbrant>	 bearND: are you cool with merging this one now: https://gerrit.wikimedia.org/r/288116
[18:00:30] <bmansurov>	 will be late a couple of mins
[18:03:20] <phuedx>	 jdlrobson: for cache misses (logged in) i think it's fine, but for anons, you have predefined buckets in the vcl, right?
[18:05:57] <dr0ptp4kt>	 joakino: updated hangout on air link: https://hangouts.google.com/hangouts/_/vjf5vvqjanf2jdm2us4eywvxkme
[18:31:36] <kevinator>	 hi mhurd , can you share the repo for your demo today?  I'd like to have a look ;-)
[18:35:46] <mhurd>	 kevinator: ya for sure! lemme get it pushed somewhere...
[18:35:48] <joakino>	 out for the day
[18:35:52] <joakino>	 bb!
[18:41:07] <bearND>	 mhurd: you could push it to the MCS repo, just start the commit msg with "WIP: "
[18:41:21] <mhurd>	 bearND: sounds good
[18:51:07] <mhurd>	 kevinator: sorry it's taking so long... something's goofy authenticating...
[18:51:19] <kevinator>	 mhurd: ok...
[18:51:23] <kevinator>	 no rush
[21:33:32] <jdlrobson>	 dr0ptp4kt: do we have a meeting?
[21:33:39] <jdlrobson>	 i think im at the wrong oom
[21:33:54] <dr0ptp4kt>	 jdlrobson: r37