[16:43:29] https://www.irccloud.com/pastebin/t0AUAFpt/ [17:05:27] just pulling out that link https://www.youtube.com/watch?v=Tb_9gOYsWJQ and wondering where everyone is?! [17:05:38] (do we do remote watch parties somewhere else now?) [17:11:39] I"m on it but just a lurker [17:12:33] * subbu is watching the youtube stream .. [17:36:31] How do you do breaking changes in event schemas or is that forbidden? [17:37:26] Reg forbidden: That brings me to the next question then. :-) [17:37:40] Events expose internal implementation details / architecture / etc. But sometimes they are just artefacts and could be a mistake. For example, events in the parsing pipeline shouldn't have been events since it exposes irrelevant internal details. i.e. events were too low-level for module contracts. What kind of guidance exists how you decide what events to expose and relevant data? [17:37:58] great talk, thanks Andrew! I have a question but I'll hold it to the end [17:38:24] although now I have a question if breaking changes are permitted across major version number boundaries [17:38:59] I can ask this :-) [17:39:35] fair enough reg data modeling qn. [17:40:40] ah, short answer "yes, with a but", long answer "no, with an if" -- got it :) [17:40:59] a thought on the modeling question is that domain driven design is the place where these data modeling questions are being answered. And Diana and Kate C are leading weekly architecture office hours where they're trying to socialize that skill set. I highly recommend it, it's a lot of fun. [17:41:53] (and I agree that getting the modeling right is an art, and the "hard" part of these systems) [17:41:55] got it .. mine was more a qn. at pushing at the boundaries of this domain. [17:42:12] to see what is known and how it is addressed / handled. [17:42:15] milimetric: are the office hours open to all? [17:42:18] yes [17:42:27] (trying to find link) [17:42:27] Q for Andrew: do you have any thoughts on how to transform event data when you don't control the producers of the events? maybe the producers are internal code of a web browser or something :) [17:42:37] Okay I'll make a note on the livestream [17:42:48] cdanis: got it [17:46:13] milimetric, are the office hours open to outside wmf though? [17:46:52] that was the question I asked [17:46:55] he said they were [17:47:19] ah, ok. yes, see that above. i only heard your announcement on stream. [17:47:41] subbu: I can't find the original announcement, but here's apaskulin to let us know! [17:48:02] apaskulin: folks were wondering about the link for the architecture team's office hours and whether it's open outside of wmf? [17:48:10] thanks so much! [17:48:39] good talk andrew! [17:49:01] ^ [17:49:44] and yes, we use kafka to purge both ATS and Varnish caches, and it is a vast improvement over the UDP multicast stuff we had going before [17:50:26] and the Network Error Logging spec is implemented in Chrome and MS Edge right now, and it's presently live on group0 right now :) [17:52:27] Hi all! I'll find out about whether the architecture office hours are open outside of WMF and share that info via the Tech Talks page. For those inside WMF, you can find the office hours on the staff calendar. You can also reach out to us at architecture@wikimedia.org if you have any questions