W3C

- DRAFT -

M&E IG monthly call: TPAC follow up

05 Nov 2019

Agenda

Attendees

Present
Chris_Needham, Shan_Huaqi, Kaz_Ashimura, Francois_Daoust, Grrett_Singer, John_Luther, Larry_Zhao, Will_Law, Pierre-Anthony_Lemieux, Rob_Smith, Steve_Morris, Peipei_Guo, Nigel_Megitt, Fuqiao_Xue
Regrets
Chair
Chris, Pierre
Scribe
kaz, cpn

Contents


<p> <em>Timestamps are in JST.</em></p>

<kaz> scribenick: kaz

<scribe> Agenda: https://lists.w3.org/Archives/Public/public-web-and-tv/2019Oct/0007.html

Chris: I'll give an overview from the f2f meeting during TPAC,
... and use some time at the end to talk about the next steps for the group,
... priorities for the upcoming period.
... Feel free to interrupt and question.

<cpn> Chris's slides

Media groups at TPAC 2019

Chris: Lots of media related groups met at TPAC this year.
... MEIG, Audio WG, Immersive Web WG, Second Screen, Web&Networks, Color on the Web, Breakouts, Media WG, TTWG
... And lots of relevant breakout sessions.
... DataCue and time marches on in HTML, Efficient audio/video processing, etc.
... Won't go into all these in detail, but the slides contain links to all the minutes for easy reference.

360 degree video

Chris: M&E IG meeting, input requested no content provider requirements
... We've had a couple of presentations in MEIG over last two years, from Fraunhofer FOKUS and IRT.
... 360 video and accessibility, lots of interest there.
... WebXR focusing on low-level input and rendering, but inefficient for video rendering.
... No consensus to extend element yet, e.g., on projection data.
... Important thing for today's call is where to follow up.
... There are two issues open in the Immersive Web CG
... Issue #55, support for 360 video
... https://github.com/immersive-web/proposals/issues/55
... Issue #40 for captioning
... https://github.com/immersive-web/proposals/issues/40
... Separately, there's now an Immersive Captions CG created, want to coordinate with them.
... Want to figure out exact need from an API viewpoint.
... TTWG members have thoughts on adding positional information to TTML.

Bullet Chatting/Danmaku

Chris: Also known as video comment overlay.
... A number of discussions at TPAC within the MEIG and the TTWG.
... What is needed to support? Possibly a delivery format?
... Maybe TTML already supports all the layout requirements.
... We need to figure out where the need is for new standards or APIs, precise synchronised accurate rendering, etc.
... There's a conference call for the Bullet Chatting CG this Friday, Nov. 8.
... We could follow up possibly in the MEIG?
... We'll discuss that during the CG call.
... Let me know if you want to join the call, I can share the details.

Huaqi: I would support the idea of creating a TF within the MEIG.

Chris: Great, thank you!

<RobSmith> https://www.w3.org/community/bullet-chatting/

Huaqi: Will you join the CG call on Friday?

Chris: Yes

Huaqi: We can talk about the detail about this topic during the CG call on Friday.

Chris: Ok, and after Friday we can report back to the MEIG.

Huaqi: Thanks

Web of Things

Chris: We didn't have enough time to talk about this during our IG meeting,
... but NHK gave demo of Web of Things for Hybridcast control API at TPAC.
... WoT IG are interested in developing media use cases.
... These could be TV as control interface for other appliances within the home,
... or media devices as discoverable/controllable devices themselves.
... WoT guys are interested in talking with us,
... offline we talked about a possible joint call, possibly December 3.

Kaz: Feedback from the WoT group,
... maybe difficult for the WoT Chair to give a presentation on Dec 3.
... So it might be better to have our own media-related discussion on Dec 3 and postpone the joint discussion.

Chris: OK

<cpn> scribenick: cpn

Kaz: I have invited WoT to provide some introductory information to share with us.

<kaz> scribenick: kaz

Chris: OK. Let's have some more offline discussion about the possible collaboration.
... In the meantime, I want to invite IG members to consider IoT / WoT use cases they may be interested in.

Media Timed Events

Chris: Discussion at TPAC of pros and cons for user agents vs application level parsing
... Common events, such as those defined in MPEG-DASH could be handled by the UA, other app-specific events in the application layer.
... I want to move this forward, get to an API specification.
... Also some good discussion about sync accuracy.
... We still have some outstanding issues from the IG Note
... There's a related discussion around a new generic TextTrackCue API,
... addresses issues around styling of cues, how to render, etc.
... At the joint TTWG/Media WG meeting, a possible TextTrack architecture was presented.
... What would be the pipeline mechanism for parsing, handling, and synchronizing cues?
... There's a general architecture for cue handling to be developed,
... for handling in-band events as well as application generated events.
... Next steps: We need some additional input around DataCue API, to make progress
... Want to look at the architecture in the context of both DataCue and generic TextTrackCue,
... There's the Sourcing In-band Tracks spec, needs updating? Need to spec how in-band cues are handled.
... DASH-IF has parallel activity, need to coordinate with them, bring their expertise.
... I propose to continue the TF for monthly conf call on Monday 18th November.
... Will send the details to the IG list.

Frame accurate seeking and rendering

Chris: A good discussion in the IG meeting, a number of use cases mentioned.
... Relates to one particular breakout as well, on efficient audio/video processing.
... Use cases such as video editing in the browser. Missing capabilties, such as seek to a specific frame, step forwards or backwards frame by frame.
... Also, someone from Amazon talking about dynamic content insertion, seamless switching between streams.
... Something we could follow up, is there something we want to contribute into the Media WG on this?
... Rendering use cases as well in addition to seeking use cases,
... graphic overlays on video, subtitle rendering sync with frame changes, synchronized playback of multiple streams, possibly across multiple devices.
... Leads into the next topic.

Professional media workflows on the Web

Chris: Media production workflows moving to the cloud, Pierre's presentation to MEIG, users expect to use web applications.
... Should we write down the gaps?

Pierre: Good summary, media editing/production is moving to the cloud,
... because users are used to using the web for everything,
... and the content itself (archive) as well is moving to the cloud.
... Expect to see media production done on the cloud side within the next a few years,
... Where are the gaps? Stereoscopic video, HDR video.
... Some are more complex, such as accurate subtitle display, frame accuracy, audio processsing.
... What's missing in the Web platform for media production?

Will: Are you thinking things like SMPTE time code notation for frame accuracy, or is it enough to get millisecond precision for seeking?

Pierre: Two issues were identified, and possibly more.
... One is the ability to seek to a specific frame, which I believe is not entirely possible yet.
... Also, today the timing precision in MSE is frame accurate, it's all floating point,
... so not a good way to precisely indicate time offsets.

Will: OK

Chris: As a next step, what do we want to do here in the MEIG,
... e.g., create a TF? A sub-group focused on a particular topic,
... looking at use cases, missing capabilities, etc., produce a document.
... Leads to proposed changes with related WGs, etc.
... How much interest is there to work on this and contribute to the topic?
... I felt from the f2f meeting there was enough interest.
... Should we make a resolution to create a TF?
... We'd need for someone to be the TF leader, or moderator.

Pierre: Maybe a good question to ask today is: do you feel there are gaps today, or not, for the professional use cases?

Garrett: I have seen some of these issues.
... Frame accuracy is important for media professionals to do editing in the browser.
... This is a valuable TF to be created, but there's overlap with other features, giving the production use case perspective would be good.
... Definitely would like to participate.

Will: Not only from a video production viewpoint, but seekability and in general,
... would be important Frame accurate splicing, client side ad insertion.

Chris: This something we, BBC, would be interested as well, we've been working on cloud based production technologies. I'd hope we could bring some input.

Pierre: Maybe what the co-chairs can do is scope this and send to the mailing list, for others not on the call.
... It's worth precisely defining what we want to do.

Chris: OK, also think about other people who may want to bring in.
... Let's follow up offline.

<scribe> ACTION: Chris+Pierre to scope the media production work and share with the IG.

Kaz: Is that kind of an initial TF Charter?

Chris: Yes

Discussion

Chris: What do you see as priorities for the IG the next months?
... There's a lot of media-related stuff in development now,
... some are exploratory, some are in incubation (WebCodecs, WebTransport),
... specs in Media WG such as capability detection and low latency.
... Feedback on those should go to the WG directly.
... If there is a use case or a requirement not currently being worked on in the Media WG, we can discuss here.
... (nothing)

Francois: You mentioned efficient media processing before.
... There was a breakout session during TPAC.
... I find it an interesting topic, it's a topic that spans multiple groups,
... each looking at a different aspect, but nobody looking from an overall perspective.
... We have a long list of use cases people might be interested in,
... some generic high-level requirements, e.g., enabling efficient media processing.
... Some parts of the solution in developmemnt, e.g., WebCodecs.
... It could be interesting to draft or scope the architectural requirements,
... e.g., a processing graph, avoiding copying memory, native or hardware GPU processing, Web Assembly.
... I'm planning to draft an architecture document, mostly for myself,
... and would propose to come back to the IG at some point for review, if there's interest.

<cpn> scribenick: cpn

Kaz: This breakout is interesting, but the topics are very broad, would it make sense to bring the results to WICG, then later to M&E IG?

Francois: It's not ready to incubate solutions yet, although some topics are in incubation now, such as Web Codecs.
... Some discussions around efficiency happening there, e.g., copying memory.
... But solving the entire space isn't in scope.
... It's things that may be of interest to media companies, related to streaming, e.g, overlays, sync.
... At this level, makes sense to discuss in the IG.
... For now this is more of a heads-up really.

<kaz> scribenick: kaz

Francois: It's useful to figure out how all the new APIs fit together.
... One thing that struck me from the breakout was the need for efficiency, today the possible frame rates are quite low,
... e.g., background removal from a webcam or the "funny hats" use case for WebRTC NV.
... At the BBC, we have a library based on WebGL for client side rendering and composition, has limitations due to inefficiencies.

Next IG call

Chris: Any thoughts on what to cover next? Could be to look at media production use cases,
... once we've looked at the scope.

Media Timed Events

Chris: What would be the best way to coordinate with the DASH-IF work on eventing?

Will: I don't have the latest status,
... but Iraj from TenCent is leading that group, he submitted a proposal to MPEG.

Chris: Now is a good time to catch up with them, work together on the web APIs.

Will: Also, WAVE wants an interoperable solution, not DASH specific, more about emsg in CMAF for event delivery.
... and have those events parsed by the browser from a SourceBuffer, which is where DataCue comes in.
... WAVE wants this to make OTT interoperable
... Clients need to understand the dispatching mechanism, e.g., real time, and where does parsing happen?
... There's some things done by W3C, DASH-IF and WAVE, needs more coordination.

Chris: I agree. I'll follow this up. Thank you.

<scribe> ACTION: Chris to contact DASH-IF about media timed events

Summary

Chris: (summarises discussion)
... We have issues in GitHub tracking these topics, feel free to comment there.

MEIG GitHub repo

[adjourned]

Summary of Action Items

[NEW] ACTION: Chris to contact DASH-IF about media timed events
[NEW] ACTION: Chris+Pierre to scope the media production work and share with the IG.
 

Summary of Resolutions

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.152 (CVS log)
$Date: 2019/11/12 03:55:28 $