Meeting minutes
Recording: https://
Slideset: https://
jib: it's "setcodecpreferences" not "setcodecrequirements" - so proposal A is acceptable
orphis: proposal b might be better (fail early)
pthatcher: not clear on send / receive codec preferences
jib: might need language to clarify "it's input to negotiation".
jib: fallback (A) should be more user friendly than failure (B)
tim panton: it's going to be pretty rare, easier to let the user who does strange things work things out than footguns. Prefer A.
orphis: A is a silent error, B is an explicit error.
RESOLUTION: Rough consensus on proposal A (fall back to "all supported codecs").
bernard: duration?
bernard: in favor of doing a PR.
youenn: this exercise needs to be done. <Not clear what is being proposed>
youenn: on the receiving side, we need the DecodedSource proposal, which doesn't exist yet.
jib: we need to solve this problem. worried about the different data semantics. Might be an alternative to use EncodedVideoChunk in webrtc transforms.
bernard: both audio and video?
hta: yes.
jib: is round-trip conversion lossless or lossy?
hta: it's lossy from RTC to EncodedVideoChunk
youennf: should consider sending encodedvideochunk + metadata to the encodedvideosource.
orphis: what happens if new fields are added in the video chunk?
hta: we will have to update both specs in tandem.
Encoder complexity control
bernard: both for audio and video?
bernard: this control exists for Opus.
youennf: don't know if the application can get enough information to use this API effectively.
hbos: much like the content-hint API.
tim panton: see an use for this when sending multiple streams of opus, for instance.
hta: we can use compute pressure as one signal
youennf: hope we have more signals than compute pressure
[[slide 26]
App control over ICE checks (restricted from previous version)
pthatcher: looks good, straightforward
sameer: ready to move to a PR.
future extensibility
pthatcher: at TPAC - asked about community group (CG) - some people interested in joining a CG that wouldn't join a WG
jib: would be good to keep it in the WG
jib: if someday we could polyfill RTPSender and RTPReceiver on top of RTPTransport, this would be something that we could get behind.
pthatcher: the hard part of polyfilling RTPReceiver is the jitter buffer.
proposal: go with #2
jib: what was wrong with what you proposed at tpac?
jib: not sure it's powerful enough to mitigate against remote control
eladalon: we can add a restricted list of events
eladalon: the TPAC proposals were clunky and unique on the web platform. This is more idiomatic.
jib: need to agree on the security level we are seeking.
youennf: look at ipad where pinch allows you to scroll & zoom at once.
youennf: could have an unified interface for both capture-wheel and set-zoom.
eladalon: don't understand why you'd want to merge these two very different user interactions.
eladalon: the demo works flawlessly with the trackpad.
jib: doesn't seem like tpac feedback has been incorporated.
eladalon: the minutes from tpac show a very different mood.
discussion....
youennf: capturewheel has a reasonable degree of consensus on the model, but has comments - will provide feedback within 1-2 weeks.
hta: capturewheel is ready to ship. we should concentrate on objections to setzoomlevel that show that there are security concerns are not mitigated.
youennf: we should consider whether reusing the direct forwarding mechanism is possible.
guidou: the setzoomlevel is interaction with an app-provided API that needs to work without a touch screen.
eladalon: note - don't look at the spec until end-of-day tomorrow.
youennf: need to file an issue with HTML - we want a stricter limit than transient activation, think we need HTML experts to chime in on this.
eladalon: the proposals #1 and #2 give stricter limitations than transient activiation (as you requested).
youennf: we want synchronous actcivation, not transient activation. html spec should define that.
eladalon: can you give me a timeline to come back with feedback?
jib / youennf: 2 weeks for feedback.
jib: limiting to click in the initial version seems good to me.