- From: Raymond Toy <rtoy@google.com>
- Date: Thu, 3 Dec 2020 14:43:22 -0800
- To: "public-audio@w3.org Group" <public-audio@w3.org>, public-audio-comgp@w3.org
- Message-ID: <CAE3TgXGKo36w1DKOsEo+vzR0hb1h9AuEEu0m11S5-_CKUqj79A@mail.gmail.com>
Attendees: Jack Schaedler, Chris Lilley, Raymond Toy
Minutes:
- Recommedatoin status
- Chris Lilley: We should probably stop accepting new issues and push
them to the next version.
- Chris: We probably want to do a CR
- Request updated CR
- Chris: There have been quite a few changes, let's do another CR
- Raymond: I agree
- Jack: Agrees
- V1
- New issues
- Audio worklet error handling
<https://github.com/WebAudio/web-audio-api/issues/2267>
- Chris: The last comment says we can close this. Can we?
- Raymond: karlt@ responded and explained things and annevk@
said we can close it.
- [Raymond to close with a short comment]
- Units & examples used in DynamicsCompressorNode are ambiguous
<https://github.com/WebAudio/web-audio-api/issues/2279>
- Raymond: Nice to have, but accept the PR.
- Chris: We really want to limit new issues
- Raymond: Can we move it to V2? And then move back to V1 if we
really want to fix it in V1?
- Chris: We can do that.
- [Raymond move to v2]
- Inconsistencies with "control thread state"
<https://github.com/WebAudio/web-audio-api/issues/2280>
- Chris: I did take a quick look and understand the issue.
- Raymond: We also have the ABSN.start issue which is somewhat
related.
- Chris: I see that now.
- Raymond: I was confused on what the ABSN.start algorithm
does. We should clean ABSN.start and handle this at the same time.
- Raymond: But I need to look to see what the issue says.
- Outstanding issues
- ABSN.start control messages
<https://github.com/WebAudio/web-audio-api/issues/2248>
- Invoking JS in the rendering loop needs a bunch of setup that
doesn't seem to be in the spec
<https://github.com/WebAudio/web-audio-api/issues/1967> and
Integration
between event-loop of AudioWorkletGlobalScope and rendering loop
<https://github.com/WebAudio/web-audio-api/issues/2008>
- PRs
- Fix #2032: ABSN extrapolates the last output
<https://github.com/WebAudio/web-audio-api/pull/2256>
- Raymond: Just want someone else to take a look
- Clarify units used in threshold & knee value calculations
<https://github.com/WebAudio/web-audio-api/pull/2273>
- Raymond: Would like Paul to take a look also since he wrote
the compressor spec.
- V2
- Issues
- NoiseGenerator
<https://github.com/WebAudio/web-audio-api-v2/issues/8>: Updates
- Raymond: Paul was to provide feedback from javascript team
about bigint.
- priority-1 issues
<https://github.com/WebAudio/web-audio-api-v2/issues?q=is%3Aopen+is%3Aissue+label%3Apriority-1>
- Render block sizes
<https://github.com/WebAudio/web-audio-api-v2/issues/13>
- Raymond: General agreement to do it, but corner cases not
discussed
- Raymond: Need to decide what values are supported and what
happens if the requested value is not supported.
- Raymond: There's also a fingerprinting issue if we report the
actual value used (for "hardware"). But this can be
obtained from an
AudioWorklet node anyway.
- Chris: Yeah.
- Raymond: Could round to supported value or throw error
- Chris: An error isn't probably helpful. Just round to
supported value and let dev decide if that's ok.
- Jack: Nodes that do FFTs could also buffer up the audio
themselves and then perform the FFT when they have enough
data available.
So it might be that the FFT size never changes
It just comes after 6.2 buffers instead of 3 or whatever.
- Raymond: That's kind of messy because you get latency in some
cases but not in others.
- Jack: Yeah, I see what you mean.
- Raymond: I think we just need to decide what is truly
required, how to expose the actual value, and what happens
if the requested
size is not supported.
- Raymond: Jack, I think you said Android support and large
sizes were the most important for you.
- Jack: Yeah, having appropriate sizes for Android where we
don't waste CPU is beneficial. Larger sizes would be good
for algorithms
that want to operate at larger sizes
- Raymond: Agreed. I'll try to ask some Android people on why
there are weird sizes.
- AudioWorkletProcessor supposed to be stateless?
<https://github.com/WebAudio/web-audio-api-v2/issues/13>
- Raymond: Link here is wrong. and I can't find the issue
- Chris: I can't either.
- Raymond: I'll try to find it and we'll take about it next
time if needed.
- Inspect CPU usage
<https://github.com/WebAudio/web-audio-api-v2/issues/40>: Updates
- Raymond: Need to decide on how accurate it needs to be, what
the units are, and how it's computed.
- Chris: Do we want instantaneous or smoothed?
- Raymond: Maybe we can add a smoothing parameter to let
developers choose the appropriate amount of smoothing.
- Jack: I think you want both
You definitely want worst case over the last N seconds or
whatever
- Jack: Something like rolling average is good to get a feel of
where you sit, and worst case is really important to know
if there's
something bad going on that's taking you over the deadline.
- Raymond: This makes sense.
- Raymond: Do we want polling or send an event if the capacity
exceeds a threshold?
- Jack: Ooh
the threshold idea is nice
- Raymond: Maybe need a low threshold?
- Chris: Yeah, we need two for hysteresis.
See you all next week!
On Wed, Dec 2, 2020 at 4:40 PM Raymond Toy <rtoy@google.com> wrote:
> Here's the agenda for our WG/CG meeting.
>
> Meeting venue https://meet.google.com/wgg-wjtd-kdu
>
> Agenda:
>
> - V1
> - New issues
> - Audio worklet error handling
> <https://github.com/WebAudio/web-audio-api/issues/2267>
> - Units & examples used in DynamicsCompressorNode are ambiguous
> <https://github.com/WebAudio/web-audio-api/issues/2279>
> - Outstanding issues
> - ABSN.start control messages
> <https://github.com/WebAudio/web-audio-api/issues/2248>
> - Invoking JS in the rendering loop needs a bunch of setup that
> doesn't seem to be in the spec
> <https://github.com/WebAudio/web-audio-api/issues/1967> and Integration
> between event-loop of AudioWorkletGlobalScope and rendering loop
> <https://github.com/WebAudio/web-audio-api/issues/2008>
> - PRs
> - Fix #2032: ABSN extrapolates the last output
> <https://github.com/WebAudio/web-audio-api/pull/2256>
> - Clarify units used in threshold & knee value calculations
> <https://github.com/WebAudio/web-audio-api/pull/2273>
> - V2
> - Issues
> - NoiseGenerator
> <https://github.com/WebAudio/web-audio-api-v2/issues/8>: Updates
> - priority-1 issues
> <https://github.com/WebAudio/web-audio-api-v2/issues?q=is%3Aopen+is%3Aissue+label%3Apriority-1>
> - Render block sizes
> <https://github.com/WebAudio/web-audio-api-v2/issues/13>
> - AudioWorkletProcessor supposed to be stateless?
> <https://github.com/WebAudio/web-audio-api-v2/issues/13>
> - Inspect CPU usage
> <https://github.com/WebAudio/web-audio-api-v2/issues/40>: Updates
> - Select audio device
> <https://github.com/WebAudio/web-audio-api-v2/issues/10>, get
> device change notification, audio config
> - Use WASM more efficiently (BYOB)
> <https://github.com/WebAudio/web-audio-api-v2/issues/4>
> - Others?
> - PRs
> - Bikeshed boilerplate
> <https://github.com/WebAudio/web-audio-api-v2/pull/95>
> - AOB
>
>
Received on Thursday, 3 December 2020 22:43:51 UTC