<scribe> scribenick: cpn

Review of recent changes

https://github.com/w3c/me-media-timed-events/pull/15

Giri: Some editorial changes from Charles Lo.
... GitHub doesn't preview the changes.

<kaz> Media Timed Events (Use cases & Requirements)

Giri: This is handled by the DASH-IF

Chris: We have a new section 4.1 for MPEG-DASH
... the idea was to separate DASH specifics from the use cases in section 3.

Giri: Two cases: static (i.e, with respect to the media itself, so in the MPD) or dynamic
... There's one outstanding PR: https://github.com/w3c/me-media-timed-events/pull/14
... Need to clarify a sentence "During that suspension interval auxiliary content such as the presentation of UI controls and media files."
... I'll check the original document.
... I think that what he means is presentation of UI controls and media files won't be possible. If we get that clarified, can we merge?

Chris: I think once that's done, it's good to merge, yes.

Giri: The other open issues contain some good discussion, e.g., https://github.com/w3c/me-media-timed-events/issues/1
... could we add tags to discuss with WICG?

Mark: I think that's a good idea.

https://github.com/w3c/me-media-timed-events/issues/3

Giri: I think Charles is happy, so can close this.

https://github.com/w3c/me-media-timed-events/issues/6

Giri: Adding use cases, let's keep open. Rob, are you willing to write something?

Rob: I can do, if I have access to the repo
... where specifically should I add?
... I'll look at other examples, and follow those.
... When is input needed?

Giri: Ahead of TPAC, if possible?

https://github.com/w3c/me-media-timed-events/issues/17

Chris: This may be an immediate response to what we're doing, "why not just use VTTCue?"

Giri: There are strong VTTCue proponents in WICG

Rob: Pre-empting TPAC discussion, is it possible to thread VMT cues through VTTCue, or is a different approach needed?

Recommendations for WICG

<kaz> Media timed events API for MPEG DASH MPD and emsg events

Chris: I posted this to the WICG Discourse forum. It's focused specifically on a subset of our TF's work, for DASH and esmg.
... Other things may not be ready to go as yet, e.g., carriage of web resources in ISO BMFF.
... Those people not been active in the TF's work, so I propose treating that separately.

Mark: I think this is fantastic. We need this to move to a proposal, let alone implementations.
... This is the right thing. Hopefully we'll get some replies, and at the IG meeting we'll be able to have some concrete follow up at TPAC

Giri: I agree.
... For TPAC, should we invite delegates from Microsoft and Google, also Apple and Mozilla? Do we know who those people are, to ask them to join the discussion?

Mark: Yes, we should invite representatives from all four.
... And I'd include Apple, as they shipped an API like this, the DataCue API.

Chris: I know who the individuals are. Not sure if they'll all be at TPAC, but we can certainly invite them.

Mark: This work may be about changes to HTML, so in scope of WHATWG.

Giri: Chris, can you contact people who are coming to TPAC?

Chris: Sure.

Next steps for the document

Giri: Chris has suggested publishing as an IG note.
... I'd like to hear what people think.

Mark: I like the idea of publishing as a draft note, wait for feedback at TPAC, then publish a final note.
... A draft note would be useful as it then has some standing, and where it's leading, in its current form. It nay encourage others to read it.

Giri: Kaz, can you advise on how to do it?

Kaz: Yes

Frame accurate synchronisation

Giri: One more thing, not in the agenda. There was a lengthy discussion on frame accurate synchronization.

https://github.com/w3c/media-and-entertainment/issues/4

Giri: Do we want to mention this in the context of our document? What would we want to say?

Chris: Some of that issue may be relevant to timed events, but not all of it.

Giri: There's been discussion in DASH-IF in the context of dispatching of the timed event.
... DASH-IF is still discussing an in-band protocol.
... Should we say something about this, even if there isn't a perfect conclusion?

Chris: Requirements from a captioning point of view?

Nigel: Threshold for acceptability of a TextTrackCue timing is 250ms to 1/10 of that.
... Chrome on Mac fires TextTrackCue events anything up to 250ms after they need to be fired.
... Firefox seems to fire within a few milliseconds, which is fine.
... We can't get accurate enough as currently implemented.
... Two use cases: Captions, which are authored for frame accuracy, at a frame rate of 25 fps.
... Too far away from that can mess up the audience experience.
... The second, is using cues to trigger playback of related audio. We've done this for audio description.
... Trigger JavaScript code to play audio. Need similar frame accuracy. If the clip doesn't play on time it can clash with the programme audio, or could be clipped (truncated).
... We can demo our implementation of a TextTrackCue with Web Audio, also at the Audio Description CG meeting.
... You definitely don't want 250 ms delay.

Mark: Are there two issues that can be handled orthogonally: accuracy in authoring and accuracy in presentation?
... Most important being frame accurate authoring.

Nigel: The actual requirements differ from what's written in the spec.
... No doubt you should be able to author to what you want.
... This GitHub issue isn't just about cue handling, but also about frame accurate seeking, which is more about control over the video than a model of the video playback.

Chris: Mark Foltz in the GitHub issue talks about rendering and the separation of the media pipeline and JS execution environment.

Rob: I think browsers would be able to get synchronisation right if they were responsible for caption rendering themselves.

Nigel: I think the main issue is that the cue onenter event is triggered late.

Giri: ATSC doesn't make normative requirements for that. Also, it's 4-5 years ago that HbbTV first alerted W3C to this problem.
... We should update the document to reflect this issue. If it's not there already, we could be more specific about captions.
... The WICG should be alerted to the more general use cases, not just timed event triggering but also caption rendering.

Chris: I suggest adding any content that relates to timed events from that discussion to our use case and requirements document.
... If there's sufficient interest, we can follow up the other parts, e.g., frame-accurate seeking, in the IG.

Rob: A couple of comments. Since I last presented to the IG, I've published the data model and syntax for WebVMT.
... That will be reviewed at the Spatial Data on the Web IG at TPAC. It outlines the details of how that data will be handled, in comparison to WebVTT.
... It may be of interest here to read that too.
... I've proposed a Wednesday breakout session about web video event support. I'm hoping to join you for the Media Timed Events session on Monday morning.

[adjourned]