[15:34:22] <Krinkle>	 Amir1: https://phabricator.wikimedia.org/T147296#2700753 brion in 2016, ref T32906
[15:34:22] <stashbot>	 T32906: Store DjVu, PDF extracted text in a structured table instead of img_metadata - https://phabricator.wikimedia.org/T32906
[15:34:45] <Krinkle>	 all the way back to 2011
[15:35:09] <Krinkle>	 I guess ten years later is still a good time to have finished it
[15:35:39] <Amir1>	 ten years is nothing in wikimedia time
[15:36:14] <Amir1>	 the videojs beta feature was deployed seven years ago and got rewritten twice before becoming default
[15:36:51] <Amir1>	 If brion claims prophecy, I will be his first follower
[15:42:22] <Krinkle>	 speaking of 2011 RFCS we haven't implemented yet, ref https://phabricator.wikimedia.org/T28741
[15:42:27] <Krinkle>	 ref T28741
[15:42:27] <stashbot>	 T28741: Migrate file tables to a modern layout (image/oldimage; file/file_revision; add primary keys) - https://phabricator.wikimedia.org/T28741
[15:43:13] <Krinkle>	 this was my first RFC, during my first in-person wikimedia or software-related event (Amsterdam 2011 hackathon)
[15:43:42] <Krinkle>	 according to this data, ignoring a volunteer's first RFC for ten years is a great retention winner 
[15:56:19] <Krinkle>	 AaronSchulz: I'm seeing about 30% of WAN getWithSet's for resourceloader-titleinfo are "hit.volatile" which seems suspicious. The TTL for this value is 1 hour, and it breaks down as 120K/min hit.good, 60K/min hit.volatile and 30K/min miss.compute. https://grafana.wikimedia.org/d/lqE4lcGWz/wanobjectcache-key-group?orgId=1&var-kClass=resourceloader_titleinfo
[15:57:00] <Krinkle>	 If the README Is correct, then hit.volatile means another thread wrote the value less than 0.1ms in the past and it is already considered expired but we take it as fresh anyway.
[15:57:11] <Krinkle>	 How can it be <0.1ms old and already expired if the TTL is 1 hour?
[17:52:17] <AaronSchulz>	 Krinkle: why is the check key touch rate so high?
[17:56:25] <Krinkle>	 AaronSchulz: right, I guess that somethign to do with it :D
[17:56:37] <Krinkle>	 some bot running wild on NS_MEDIAWIKI?
[17:58:27] <AaronSchulz>	 why does RL\WikiModule::invalidateModuleCache get called on null edits?
[17:58:52] <AaronSchulz>	 I don't see an $this->options['changed'] check
[17:59:10] <AaronSchulz>	 so any null edit to a user css/js page would purge all the RL modules on a wiki
[18:06:04] <Krinkle>	 AaronSchulz: well, we use page_touched not page_latest, afaik that's bumped on null edit, right?
[18:06:32] <Krinkle>	 if it's only bumped on proper purge and links update, we could skip it
[18:07:04] <Krinkle>	 but at least there is the use case of fixing gadgets cache stale by doing a purge on the JS page, that should work and is why we use page_touched
[18:33:40] <AaronSchulz>	 yes, page_touched is updated for null edits