[00:35:56] <jackmcbarn>	 on the wmf cluster, how are locally customized interface messages cached?
[02:11:48] <Betacommand>	 jackmcbarn: why do you want to know?
[02:12:29] <jackmcbarn>	 Betacommand: i care whether or not it has to make a request over the network for them
[02:13:00] <Betacommand>	 jackmcbarn: depends on if its in the varnish cache or not
[02:13:17] <jackmcbarn>	 i thought varnish only caches whole pages
[02:13:42] <Betacommand>	 jackmcbarn: correct, but if enough people use that language its in varnish
[02:14:07] <jackmcbarn>	 for my purpose, the whole page doesn't matter
[02:15:27] <Betacommand>	 then you have the parser cache, object cache and l10n_cache
[02:15:50] <Betacommand>	 depending on exactly what your looking for
[02:17:23] <Betacommand>	 jackmcbarn: without more context of what your looking for its hard to give more guidance 
[02:17:40] <jackmcbarn>	 Betacommand: in this case, it's enwiki's MediaWiki:Spam-blacklist
[02:17:43] <jackmcbarn>	 which is huge
[02:18:05] <jackmcbarn>	 i'm currently writing code to locally cache the global blacklist to save on network, but i'm not sure whether i should do the same for the "local" one
[02:18:10] <Betacommand>	 jackmcbarn: thats not an interface message
[02:18:14] <jackmcbarn>	 since "local" is still db, but i know some messages are cached
[02:18:20] <jackmcbarn>	 the code uses it as if it were one
[02:19:12] <jackmcbarn>	 it calls wfMessage( 'spam-blacklist' )->inContentLanguage()->plain() to get the content of it
[02:20:02] <Betacommand>	 jackmcbarn: its probably then pulling from the l10n cache
[02:20:35] <jackmcbarn>	 does the l10n cache hold the on-wiki version of messages then?
[02:20:48] <Betacommand>	 yeah
[02:20:58] <jackmcbarn>	 ok, good
[02:21:00] <Betacommand>	 or the default if not modified
[02:21:36] <Betacommand>	 jackmcbarn: If you use the correct backlist config setting its cached too
[02:24:09] <Betacommand>	 jackmcbarn: I use the global blacklist on my local wiki and it seems to cache the meta page, and it keeps the quires to a minim 
[02:24:37] <jackmcbarn>	 i know it caches that part; i'm building a second layer of caching for it
[02:24:55] <Betacommand>	 why?
[02:25:55] <jackmcbarn>	 because a few months ago, i tried to add some stuff that ran the blacklist per-parse, and it overwhelmed our memcached
[02:27:07] <Betacommand>	 how big is your wiki?
[02:27:42] <Betacommand>	 and what are you trying to do on each parse?
[02:27:55] <jackmcbarn>	 this is for the wmf cluster
[02:28:23] <jackmcbarn>	 and i'm trying to make it check on parse instead of on edit so it's not so trivial to circumvent
[02:28:30] <jackmcbarn>	 this second cache layer was tim's idea
[02:30:19] <Betacommand>	 jackmcbarn: The actual cases of people circumventing the blacklist is rather insignificant
[02:30:40] <Betacommand>	 it involves templates normally
[02:30:43] <jackmcbarn>	 there's other features planned too, but they all depend on running it per-parse
[02:34:28] <Betacommand>	 jackmcbarn: is the thrashing persistent or does it drop off after a period?
[02:34:39] <jackmcbarn>	 Betacommand: it was persistent
[02:35:01] <Betacommand>	 jackmcbarn: how long did you let it continue for?
[02:35:22] <jackmcbarn>	 i think it was up for about a day before someone from ops reverted it
[02:38:09] <Betacommand>	 jackmcbarn: have you thought of comparing the ellinks before and after the parse and only running the tests if they differ?
[02:38:30] <jackmcbarn>	 that doesn't work for what i was doing
[02:40:06] <Betacommand>	 why not?
[02:40:19] <Zhaofeng_Li>	 Was there any hiccup on Tool Labs earlier? https://en.wikipedia.org/w/index.php?title=User_talk:Zhaofeng_Li&diff=prev&oldid=627837827
[02:40:45] <Zhaofeng_Li>	 I'm surprised to see that ^
[02:40:56] <Betacommand>	 Zhaofeng_Li: there where some DNS issues
[02:41:34] <Zhaofeng_Li>	 Okay.
[02:41:35] <jackmcbarn>	 if the page contains any blacklisted links, even if not new, it hid them
[02:42:12] <Betacommand>	 jackmcbarn: Uh, you will piss of a LOT of people if that happens
[02:44:10] <Betacommand>	 jackmcbarn: if you have the previous link data just include a Boolean for blacklisted or not
[02:44:38] <jackmcbarn>	 that wouldn't react well with new links being added to the blacklist
[02:50:02] <Betacommand>	 jackmcbarn: if you rely on edit/purge only for a full refresh, and leave the background reparses using the diff links it should create a nice balance for what your looking for
[02:50:44] <Betacommand>	 jackmcbarn: why are you interested in the SBL?
[07:17:05] <comets>	 aww, why was User ID removed from centralauth? :(
[07:21:49] <RD>	 comets: Who wants to know!
[07:22:26] <comets>	 it helped me keep track of the number of ppl on wikimedia :P
[07:25:09] <comets>	 removed a few hours ago..was there earlier today..
[19:09:27] <natuurtje>	 MrAjedrez
[19:09:55] <MrAjedrez>	 Hi, natuurtje 
[21:20:10] <Nemo_bis>	 imagine a world where autologin on sisterprojects works
[21:21:52] <FlorianSW|away>	 @Nemo_bis: works for me :)
[21:22:50] <Nemo_bis>	 FlorianSW|away: I'm sure Elsevier says the same of their journals :D
[21:23:26] <FlorianSW|away>	 Nemo_bis: :D
[22:07:14] <Nemo_bis>	 hmm, so the automerge upon login should now be live on all wikis, right?
[22:08:34] <Nemo_bis>	 should https://gerrit.wikimedia.org/r/#/c/162971/
[22:10:43] <sjoerddebruin>	 oh nice