[00:30:14] ping me on irc with the link, that's probably easiest [08:44:11] ori: draft up at https://docs.google.com/document/d/150SO1AR7ubs-Pk1-a59sYZsCa2KtkH49t53bcxCuHlU/edit [08:44:39] I'm looking to shorten it still, but I won't add anything new at this point. Thanks! [08:54:05] https://calendar.perfplanet.com/2020/abandonment/ very interesting data shared by Nicolás [09:31:24] Oh I missed that [10:10:22] https://www.bleepingcomputer.com/news/google/google-chrome-is-getting-a-real-time-webpage-performance-overlay/ [18:21:09] Hi perf team. [18:21:16] rather random question: how expensive from performance perspective it is to generate a UUIDv1 in mediawiki? Say we wanted to generate one for each ParserOutput we save in the ParserCache (only upon save, not upon get) - does your gut say it's a terrible idea? [18:46:00] If it can be done post-send, then it won't matter as much. [18:46:10] I'm curious from a non-perf perspective what it would be used for, though :) [19:00:59] Krinkle: we are slowly moving restbase features into core so that VE could work just for core [19:01:31] in VE, we need to stash the ParserOutput under a unique key so that after the edit we can submit it to parsoid for conversion [19:01:56] the key should be very unique, since a collision/mismatch will result in a dirty diff [19:02:34] also parsoid parses are not reproducible, so we can't just key on [19:03:06] in ParserOutput we have cacheTime already, but it's TS_MW with 1 second resolution.. too low resolution. [19:03:42] Hm.. not sure I get how the UUID helps, though. We don't store multiple versions of the same page/rev/opts combination afaik, so if we still have the same version in cache, we can use it, and if we don't have it/ it dissappeared, then we can't. How does an ID change it? [19:04:25] before giving VE the html, we stash it temporarily. The uuid is the key for that stashed entry [19:05:34] it would be good to have ID attached to the PArserOutput in parser cache - then if we need to stash it multiple times, all stashed entries have the same id [19:06:25] Ah so it's for de-duping the stash entries [19:07:18] attaching it to ParserOutput - yeah, for de-duping. [19:07:41] also we give the ID to VE and to other clients to submit back later [19:07:50] worth documentin inline when we add this in core. so that it can be easily revisited in the future. [19:07:52] and for b/c it would e nice to keep it being UUIDv1 [19:08:18] but yeah, if it's saved post-send (is it?) should be harmless. [19:08:28] yeah. For now I will just do everything in the new endpoints we are making for it. Won't propagate to too deep into core yet [19:08:42] this is another reason, though, to push our purge refactor so that we do less of this in the critical path on GET after expired cache [19:10:42] thank you!