[02:14:31] <Krinkle>	 TimStarling: about T390929, while I wait for an answer from SRE, do you have thoughts about risks around adding Vary? anything to look out for that would warrant a rollout other than riding the train? 
[02:14:32] <stashbot>	 T390929: MobileFrontend should declare "X-Subdomain" variance via "Vary" response header - https://phabricator.wikimedia.org/T390929
[02:14:52] <Krinkle>	 (and thus potentially a less trivial implementation)
[02:19:56] <TimStarling>	 I'm not familiar with how Vary is implemented in Varnish and ATS
[02:20:57] <TimStarling>	 it's possible that when a backend response is received with a new Vary header, the existing cache for that URL will be invalidated
[02:21:32] <TimStarling>	 so the risk would be a rise in backend requests as the cache is rebuilt
[02:29:49] <Krinkle>	 Thanks, I hadn't thought of that possibility yet.
[03:14:32] <TimStarling>	 Krinkle: what do you think about deprecating X-Subdomain at some point?
[03:15:37] <TimStarling>	 I suggested it in the last paragraph of https://phabricator.wikimedia.org/T390924#10926212
[03:25:47] <Krinkle>	 TimStarling: sure, naming can be improved, it is effectively that already. MF doesn't use the value. It treats it as a boolean.
[03:26:15] <Krinkle>	 the name is hardcoded in various places. Once that is fixed, we could rename it. MF mostly has it configurable already.
[03:26:34] <Krinkle>	 I say mostly because in reviewing it for these tasks I spotted a place where it hardcodes it
[03:26:58] <Krinkle>	 X-Ismobile: 1 would be my suggestion, to minimise the change. 
[03:27:28] <Krinkle>	 fixme at https://gerrit.wikimedia.org/r/c/mediawiki/extensions/MobileFrontend/+/1161099/
[03:28:20] <Krinkle>	 also, MF currently evaluates this header currently only if MobileCallback is set and returns a non-equal value (i.e. the site has an m-dot domain). Which I'm fixing at T397387, which is a step toward that direction. 
[03:28:21] <stashbot>	 T397387: Prepare MobileFrontend to support enabling $wgMFMobileHeader without an m-dot subdomain - https://phabricator.wikimedia.org/T397387
[03:28:30] <Krinkle>	 The name and docs suggest it work this way already, but it doesn't.
[03:53:59] <Krinkle>	 OK. I'll leave the patches like that for today, they're ready for review. Signing off now.