17:06:35 <RRSAgent> RRSAgent has joined #webperf
17:06:35 <RRSAgent> logging to http://www.w3.org/2017/06/14-webperf-irc
17:06:41 <Zakim> Zakim has joined #webperf
17:13:20 <igrigorik> f2f agenda: https://docs.google.com/document/d/1fcWpKZhJdrmBlpYUB2wIx42svo9zqojoJMZg8aY7E7Y/edit#
17:13:40 <igrigorik> some suggestions for topics at the bottom of the doc
17:14:10 <yoav> I'll try to rejoin
17:14:19 <NicJ> NicJ has joined #webperf
17:14:38 <igrigorik> Agenda: https://docs.google.com/document/d/10dz_7QM5XCNsGeI63R864lF9gFqlqQD37B4q8Q46LMM/edit#
17:15:09 <igrigorik> HR-Time: prompt to unload test
17:15:32 <igrigorik> xiaoqian: the test seems to work in FF, but does not yet work in Chrome
17:15:51 <igrigorik> ... need to try in Edge and Safari
17:16:05 <xiaoqian> https://www.w3.org/2017/04/webperf-test/unload_manual.html
17:17:53 <igrigorik> re, worker reference: we'll use whatwg
17:18:10 <igrigorik> Next up, feedback on PerformanceObserver + Timeline
17:18:17 <igrigorik> https://github.com/w3c/hr-time/issues/32
17:18:24 <igrigorik> https://github.com/w3c/hr-time/issues/46
17:18:38 <igrigorik> NicJ: are perf observer and timeline buffers different?
17:19:55 <igrigorik> ... not clear on relationship between, say, RT buffer and performance timeline buffer?
17:20:07 <igrigorik> ... how does it compare to ServerTiming entries
17:24:25 <igrigorik> we have some undefined behaviors: we defined new buffering in "global performance timeline", but haven't defined how the global buffer interacts with RT buffer, etc.
17:24:59 <igrigorik> NicJ: scripts like boomerang are not guaranteed to load before end of onload
17:25:09 <igrigorik> ... we can use method queue as a workaround
17:26:56 <igrigorik> second issue: dealing with parent iframe is easy, dealing with children (same-origin) iframes is harder
17:27:12 <igrigorik> ... the challenge is you have to register the perf observer before onload of the iframe
17:27:43 <igrigorik> ... there are race conditions that are hard to deal with
17:28:02 <igrigorik> ... LongTasks has an interesting model where it bubbles up all child-iframe events as well
17:28:15 <igrigorik> ... one idea is to enable perf observer to do the same?
17:29:02 <igrigorik> ... we'd like to use PerfObserver more, but the iframes make it a challenge
17:30:56 <igrigorik> yoav: having consistent buffering behavior is good for developers
17:32:30 <igrigorik> AI's to followup on:
17:32:52 <igrigorik> - define global perf timeline buffer: how does it interact?
17:33:09 <igrigorik> - iframes are tricky with PerfObserver, can we do something here?
17:36:05 <igrigorik> - debugging: bad thing happened, how can we support this case?
17:38:03 <igrigorik> AI: nick to split last question into separate discussion
17:38:24 <igrigorik> Server Timing: https://github.com/w3c/server-timing/issues/14
17:39:41 <igrigorik> ... it's hard to line up RT and ST entries?
17:40:29 <igrigorik> shubhie: it's not a unique problem.. e.g. inflight requests, slow frames and long tasks
17:40:56 <igrigorik> ... perhaps there is some generic solutions?
17:41:23 <igrigorik> yoav: for dependencies, the idea I explored is a unique ID for each entry, so you can create relationships between entries
17:41:35 <igrigorik> ... similarly, one could use that to link other types
17:42:47 <igrigorik> ... we could have a generic mechanism?
17:43:39 <igrigorik> Nathan: we have a problem where we can't match Fetch requests with resource timing
17:44:46 <igrigorik> what's the argument for having RT and Server Timing separate?
17:45:03 <igrigorik> Charles: perhaps in the future if you have a lot of different extensions.. it paves the way for separate APIs that are stitchable
17:46:17 <igrigorik> Shubhie: we'll need this mechanism anyway
17:46:44 <igrigorik> Charles: one worry with pointer idea is that these might not exist yet..
17:48:10 <igrigorik> AI: discuss next week at F2F
17:48:30 <igrigorik> NicJ: PerfObserver + async callback
17:49:01 <igrigorik> ... you pass in buffered: true, there is no signal that the are no entries
17:49:41 <igrigorik> ... https://github.com/w3c/performance-timeline/issues/74
17:51:29 <igrigorik> ^ would address the case
17:53:43 <igrigorik> Preload
17:53:44 <igrigorik> https://github.com/w3c/preload/issues/99
17:54:01 <igrigorik> yoav: fastly presented their h2 push at HTTP workshop in London.. rel=preload came up
17:54:24 <igrigorik> ... feedback on no-push was not very positive
17:54:37 <igrigorik> ... h2o implemented opposite signal 'h2-push-only'
17:54:56 <igrigorik> ... we opened an issue to explore if explicit push signal would be better
17:55:29 <igrigorik> ... not clear on use cases for the opposite signal, asked for more details on the bug
17:57:27 <igrigorik> ... let's wait for the use case
17:58:04 <igrigorik> yoav: re, f2f agenda..
17:59:29 <igrigorik> ... Wenbo (Google) asked about compression API. Many Google properties interested in compressing upstream data.
17:59:44 <igrigorik> ... in theory, not hard to implement for gzip / brotli?
18:00:05 <igrigorik> Nathan: we're interested in this as well, e.g. for perf logging -- we implement our own hacky compression
18:00:15 <igrigorik> ... lots of use cases for it
18:01:29 <igrigorik> ... for FB including gzip library is too much
18:03:24 <igrigorik> let's discuss at F2F
18:04:10 <igrigorik> NicJ: will try to get numbers for implications of size of library, transferred bytes
18:04:57 <igrigorik> Yoav: SpeedCurve is hosting an event the day of f2f
18:05:35 <igrigorik> ... please RSVP
18:05:45 <plh> plh has joined #webperf
18:07:00 <xiaoqian> RRSAgent, make minutes
18:07:00 <RRSAgent> I have made the request to generate http://www.w3.org/2017/06/14-webperf-minutes.html xiaoqian
18:07:06 <xiaoqian> RRSAgent, make log public
18:09:47 <n8s> n8s has joined #webperf
18:15:14 <yoav> yoav has joined #webperf
18:31:28 <n8s> n8s has joined #webperf
18:51:28 <n8s> n8s has joined #webperf
19:54:52 <cristi> cristi has joined #webperf
20:28:43 <Zakim> Zakim has left #webperf