See also: IRC log
<rtoyg_m_> present rtoyg_m
<hoch> +present hoch
<rtoyg_m_> +present rtoyg_m
<joe> padenot we see you twice in the WebEx list but neither one is connected
also you can try connecting vie WebRTC in Firefox, now
CSAIL, 32 Vassar St. room D463 Star
Closest Red line T stop is Kendal MIT
<joe> https://github.com/WebAudio/web-audio-api/pulls
<scribe> scribenick: chrisl
padenot_: merged a number of PR this week
<joe> https://github.com/WebAudio/web-audio-api/pull/1204
padenot_: lgtm
<ghaudiobot> [web-audio-api] joeberkovitz closed pull request #1204: Disallow empty string as AWP registration key. (gh-pages...1172-no-empty-AWP-names) https://github.com/WebAudio/web-audio-api/pull/1204
joe: merging now
<joe> https://github.com/WebAudio/web-audio-api/pull/1201
<joe> https://github.com/WebAudio/web-audio-api/pull/1200
<ghaudiobot> [web-audio-api] rtoy closed pull request #1201: Fix #1157: Add attributes to Media{Element,Stream}AudioSourceNode (gh-pages...1157-add-attributes-again-for-measn) https://github.com/WebAudio/web-audio-api/pull/1201
<ghaudiobot> [web-audio-api] joeberkovitz closed pull request #1200: Reword treatment of starting time for AudioScheduledSourceNode (gh-pages...915-subsample-accurate-start) https://github.com/WebAudio/web-audio-api/pull/1200
<joe> https://github.com/WebAudio/web-audio-api/pull/1170
<joe> https://github.com/WebAudio/web-audio-api/pull/1065
<joe> https://github.com/WebAudio/web-audio-api/issues?q=is%3Aopen+is%3Aissue+no%3Amilestone
<joe> https://github.com/WebAudio/web-audio-api/issues/1205
joe: polishing factory methods not very important, lets close it
<joe> https://github.com/WebAudio/web-audio-api/issues/1203
ray: adds nothing, lets remove it
joe: yes, but v1 blocker?
ray: no
<joe> https://github.com/WebAudio/web-audio-api/issues/1202
joe: ok, no way to stop routing audio to the node. How big a use case where a nde switches from serving to not serving the graph
padenot_: have had people requesting this feature, told them it can't be done
joe: why?
padenot_: different playback
pipelines, with normal media and a single media element and msc
to get decrypted media
... didn't want to have two media elements
joe: sounds like a v.next
<joe> https://github.com/WebAudio/web-audio-api/issues/1199
roy: hongchan reminded me convolver node can be set more than once
hoch: discussed at tpac 3 years ago. cwilso argued against making it unchangeable. Wanted to change impulse response on the fly
padenot_: too late to change
now
... on ABSN can't set multiple times but convolver you
can
... should initialise it
joe: think this is an error on my
part, adding settable once to convolver and didn't do it
... wait, did and it got taken out
ray: bad merge maybe
joe: will look at this and fix it
ray: are we keeping convolver settable multiple times or not?
<joe> https://github.com/WebAudio/web-audio-api/issues/667
<joe> https://github.com/WebAudio/web-audio-api/issues/667#issuecomment-249143280
joe: will look back on what we decided
jerry: we had a PR and it got dropped?
joe: yes. wait, seeing that language in the spec!
joe: OK we need to say it is initialised to false. And spec text on convolver stays as it is, settable once
<joe> https://github.com/WebAudio/web-audio-api/issues/1194
hoch: domenic asked me to fix the
IDL because handler being part of the node does not make sense,
can't handle the event
... anne is asking why it is impossible to be an event
target
... want to minimise functionality of descending data
... cause a lot of GC
... discussion ongoing
... anne and domenic are looking more at the whole web platform
while we want the audio worklet to work well for its intended
purpose
joe: hooking it up to worklet causes this uncertainty
padenot_: web platform implies
some things that are complicated and unclear. audio api can't
use the normal event loop
... we can do our thing and have patterns that resemble the
normal web platform. but not exactly the same
... issue is to not have an event handler
hoch: they want to expose event target on worklet
padenot_: common infrastructure is good but they don't understand at first the specific needs of audio
joe: what about the very last comment on GC?
padenot_: yes, transferable
solves this, you bet a pointer instead.
... you get the buffer inside the worklet and treat it how you
want. You can let it GC (not smart) or send in back for GC on
the main thread
joe: okay, let this active thread continue for a bit
ray: v1 blocker?
padenot_: yes. important to make
sure we address our needed use cases and efficiently access
large buffers is critical
... worklet has always been a blocker, necessary to resolve,
implement and ship for v1
... anne's suggestion might be worth it
... get transferable for free
define a normal event loop, get the buffer before rendering. ordering important
ray: we are implementing now and this changes things a lot
padenot_: pretty lightweight, bu requires some sort of event loop
joe: we should not let other groups uncertainties halt our progress. worry that may be the case here
padenot_: agreed
... need to chat more and explain more
<joe> https://github.com/WebAudio/web-audio-api/issues/1193
hoch: this one is easy, see summary at end of thread
joe: is this a v1 blocker?
hoch: yes
<joe> https://github.com/WebAudio/web-audio-api/issues/1188
chris: suggest closing the 'do nothing' one. not useful
padenot_: was wanting to close it after discussion
joe: lets close
jerry: there are workarounds, this is just for linking different status
padenot_: big audio apps just use agin nodes everywhere
<joe> https://github.com/WebAudio/web-audio-api/issues/1182
ray: can't create a type error in chrome. should be handled at the binding layer, like range errors
padenot_: how do you do that at the binding layer, in IDL? or is it spec text?
ray: not sure
... these are enums except for range error etc which are not
the same type
... we already throw range error
padenot_: like range errors, more informative, it is the right type but out of range
ray: will follow up internally
8 open blockers, want to be zero by f2f in June
padenot_: after fixing constructor I can get to the dynamics compressor
chris: lots of little PR should help
padenot_: yes it will be fine
ray: IDL is frozen after the constructor stuff, we can update to latest respec after that and it will fix it
jerry: ordering is similar but not identical. wav format extensible is a very old spec. creates some problems
padenot_: we need to pick something
joe: we did, we picked smpte
jerry: sorting out with media pipeline folks
padenot_: we use smpte in Gecko,
not set in stone, was unspecced so we went for that
... remap channels soconsistent across all codecs and
containers
joe: next call in 2 weeks
(adjourned)
This is scribe.perl Revision: 1.152 of Date: 2017/02/06 11:04:15 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00) Present: Joe Jerry Paul Hongchan Ray Chris Found ScribeNick: chrisl Inferring Scribes: chrisl Found Date: 13 Apr 2017 Guessing minutes URL: http://www.w3.org/2017/04/13-audio-minutes.html People with action items:[End of scribe.perl diagnostic output]