HTML Media Task Force Teleconference

09 Jul 2013


See also: IRC log


Paul Cotton
Mark Watson


<trackbot> Date: 09 July 2013

<paulc> Good morning

<paulc> I will need a scribe today.

<adrianba> ScribeNick: markw

<adrianba> Scribe: Mark Watson

<paulc> Agenda: http://lists.w3.org/Archives/Public/public-html-media/2013Jul/0008.html

<Michael_Thornburgh> you're typing aaac instead of aacc

<paulc> Usual pattern: "Topic: xxxxx"

Previous meeting minutes

paulc: Propose to note previous minutes

Review of action items and issues

paulc: Embedded in agenda

MSE status and bugs

<paulc> Latest Ed draft: http://dvcs.w3.org/hg/html-media/raw-file/tip/media-source/media-source.html

<paulc> MSE bugs: http://tinyurl.com/6pdnzej

Pre-Last Call bugs

<paulc> Bug 22148 - Request that we reconsider adding jitter to video quality metrics


paulc: This is open, assigned to Aaron

acolwell: This this is answered. I will try to get this implemented this week.

paulc: If this is the only open one, we could go to LC before implementing it

Bug 22137 - changes in number of audio tracks during advert insertion


<adrianba> ACTION-21?

<trackbot> ACTION-21 -- Adrian Bateman to work with Jerry to review bug 22137 and provide feedback -- due 2013-07-02 -- PENDINGREVIEW

<trackbot> http://www.w3.org/html/wg/media/track/actions/21

<paulc> See latest comment at https://www.w3.org/Bugs/Public/show_bug.cgi?id=22137#c5

jdsmith: We had some discussion and comments are in the bug. Some aspects tricky. Trade-off between throwing an error if track numbers are reduced vs trying to play.
... Trying to play is a good goal
... Local language might be preferred for example

adrianba: Aaron's compromise proposal was what the original bug was asking for. In summary the difference between what the current spec says (prohibit track changes) and coming up with heuristics for deciding on which track to continue with when available tracks change
... not obvious we have a clear answer
... in out media engine we look at languages, formats and pick default track. Other media engines might do something different. Getting interop might be hard.
... Is this something we really need to solve now ?
... If yes, how precisely should heuristics be defined and does it need to be repeatable ?

Michael_Thornburgh: I'll add a note today to eb bug about the main usecase: separate video and audio tracks coming in separately and then switch to an add which is multiplexed
... so you are going from 0 to 1 track of a particular type or 1 to 0

BobLund: Issues adrianba raised are independent of MSE and apply to any media resource where the set of tracks changes. There's a resolved bug in the HTML5 spec on this.

<paulc> What is the HTML5 bug number?

BobLund: There is some ambiguity about how applications should respond, but it's not just an MSE issue
... MSE issue is whether we want to allow those track changes at all - a separate issue from how you deal with them when the happen
... Should just discuss whether we allow track changes in MSE

<paulc> Is this a case where the spec should say the choice of text track is "implementation-defined"?

paulc: adrianba, is this a case where we should just say "implementation defined" ?
... if we did this and got into Last Call it could be a source of objections

acolwell: First on Michael_Thornburgh's use case, those transitions would actually introduce new tracks
... if one of the source buffers suddenly had an extra track, then that should introduce a new track in HTML5 and you ought to actually select that new track
... not clear that UA could automatically detect that it should switch just based on the absence of data in the old source buffer and the appearance of the new track

<BobLund> http://www.w3.org/TR/2012/CR-html5-20121217/embedded-content-0.html#audiotracklist, http://www.w3.org/TR/2012/CR-html5-20121217/embedded-content-0.html#texttracklist

acolwell: Second, on BobLund's comment, I think HTML has language about tracks coming and going which is fine but the concern is whether the media engines can actually support that and what bar do we want to set for minimal client implementations
... seems to me it is more in the scope of MSE than HTML5

pal: There is already in HTML5 a means of selecting which track should be played, correct ?

acolwell: yes

pal: so for this use-case with an ad insertion the app can select which track will start playing

acollwell: when they append the new data they should get a notification of the new track on the HTMLMediaElement and the app could select then. But there is no implicit way for the track selection to happen automatically

pal: would seem dangerous to have a track selected automatically
... another thing is whether we should impose constraints on track switching
... perhaps requires multiple decoders etc.

acolwell: current text does not allow track count or codecs to change exactly for that reason
... my comment was just to define minimal behaviour so that if the track counts do change the behaviour is deterministic, but don't require the UA to actually handle the new tracks

BobLund: links above are to the HTML IDL that provide add/remove track capability
... app can select the tracks - no ambiguity there - I agree with acolwell that the issue for MSE is whether or not to support these track changes
... if we do support them the UA behaviour is clear in HTML

<paulc> http://www.w3.org/TR/2012/CR-html5-20121217/embedded-content-0.html#audiotracklist-and-videotracklist-objects

adrianba: in response to BobLund, HTML5 is clear what event should be fired in response to track add/remove.
... what's missing is what to do if the track that you are playing disappears. Should you enable another track automatically ?
... and how do you decide which one ?
... alternatives are to give an error and stop playing or to continue playing
... ideally we would have the ability to schedule a track change at some time in the future
... so for example when appending the new track you could let the media element know that you will want to change at some future point
... might not be able to do this right away, but seems like a long-term goal
... propose converting this to an HTML5 bug

<BobLund> +1 to moving this to an HTML5 bug

adrianba: proposing future scheduling of track changes

acolwell: agree with adrianba that providing information about how to switch tracks at a particular point in the timeline
... immediate problem is whether to trigger an error (which is what is says now) or whether to change that to say it's allowed but may not be exposed by the UA
... so then this new future spec can be used to control the changes

<paulc> Mark: A simple suggestion is to possibly mark one of the tracks as the default if one of the tracks goes away.

<paulc> Mark: If the audio track goes away maybe the UA will stall and we don't want the UA arbirtarily picking tracks.

markw: perhaps allow a track to be flagged from the MSE side as the default track to be used if the currently playing track ends

<paulc> Aaron: There is currently no way for the applications to know the track ids.

acolwell: that could be tricky because There is currently no way for the applications to know the track ids

Michael_Thornburgh: provding information for track switches at append time would unify the append time behaviour with the rendering behavior
... would also solve the stall problem when there were suddenly zero tracks
... I'll add this to my comment on the bug

pal: acolwell, when you said the MSE had no concept of trackid, what did you mean

acolwell: track selection is abstracted away from track ids in the media itself. There is just a track list and you can select from that. You can't use track id on that side.

pal: There are attributes like kind, which you can use, but there's no id on the media element side, right ?

acolwell: yes, there's no id which is guaranteed to be the id from the container

pal: ?

acolwell: may be benefits of exposing the track ids, but that's not this particular issue

pal: application should be able to select which track to play from what it has inserted and wants to play the French track or track id 5

<adrianba> We had MSE bugs for this issue

acolwell: you can't select by track id but you can select by language etc.
... underlying container id is not exposed. it can change

adrianba: this topic is the subject of MSE bugs, we discussed a long time ago that mapping from track id to HTML track lists is tricky. Thought we had resolved that
... should check in time as there's another issue
... summary: propose to move this to be solved in HTML5 in a broader way

paulc: in 5.0, the CR ?

adrianba: no, 5.1
... feature request

acolwell: would that HTML5 proposal remove the current resriction in MSE ?

<paulc> Move to HTML5 -> create a new 5.1 bug that references this bug

adrianba: right, I'm proposing we make no change to MSE now and potentially change that in future

paulc: this means creating a new HTML5 bug referring to this one
... is there consensus today to do that ?
... not today

<adrianba> Bug 22136 - Inband Storage for SPS/PPS in ISO BMFF

<adrianba> https://www.w3.org/Bugs/Public/show_bug.cgi?id=22136

<adrianba> ACTION-22?

<trackbot> ACTION-22 -- Adrian Bateman to work with jerry to confirm the proposal in 22136 is acceptable -- due 2013-07-02 -- PENDINGREVIEW

<trackbot> http://www.w3.org/html/wg/media/track/actions/22

jdsmith: There's an MPEG DASH specification that requires inband SPS/PPS support
... for ISO BMFF part 15. Currently in draft state
... hard to require support when it is not yet stable

paulc: what is the exact status
... ?

jdsmith: don't know exact status

paulc: we should find that out first


DIS ballot, close of voting

adrianba: can we close ACTION-21 and ACTION-22 ?
... if we identify status of ISO spec as unstable, then what ?

paulc: markw identified status as DIS, which is pretty stable

The status is at: http://www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=62115

Getting the actual document is a different matter

paulc: looks like it when into DIS in April
... suggest adrianba looks at how stable this is

Note this is just from searching for 14496-15 amendment 2 which is HEVC support. I am not *certain* this contains the SPS/PPS thing

though that would make sense

<adrianba> ACTION-21: need to confirm exact status of the proposal for SPS/PPS

<trackbot> Notes added to ACTION-22 Work with Jerry to review bug 22137 and provide feedback.

Bug 22117 - Add a conformance section

paulc: asked Cyril what he want to do about this item
... waiting for response

Bug 22112 - Merge appendBuffer and appendStream text


paulc: This is resolved, fixed

<paulc> https://www.w3.org/Bugs/Public/show_bug.cgi?id=22112 is resolved fixed

paulc: those were the 5 bugs blocking LC. One is resolved. The first one Aaron will implement in the next two weeks
... 22137 and 22136 need more work and we have action items for those
... 22117 is on Cryil

Candidate Last Call bugs


(scribe missed that)


paulc: Aaron offered to respond at the Jul 2 meeting

acolwell: yep, will do that

Any other business

acolwell: fixed the statistics to return an object, instead of a method
... this applies to buffered attribute as well
... it returns a new object every time
... which was identified as a problem for the video playback quality
... curious whether we should also change buffered

adrianba: stick with what we have got as it's the same as the HTML5 buffered attribute

acolwell: ok, fine

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.138 (CVS log)
$Date: 2013-07-09 16:02:18 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.138  of Date: 2013-04-25 13:59:11  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/objectiosn/objections/
Succeeded: s/woudl/would/
Succeeded: s/ACTION-21/ACTION-22/
Succeeded: s/tha/that/
Found ScribeNick: markw
Found Scribe: Mark Watson

WARNING: No "Present: ... " found!
Possibly Present: ACTION-21 Aaron Aaron_Colwell BobLund Mark Mark_Watson MartinSoukup Michael_Thornburgh Microsoft P21 ScribeNick aa aaaa aabb aacc acollwell acolwell adrianba davide glenn html-media https jdsmith joined markw pal paulc pladd trackbot wseltzer
You can indicate people for the Present list like this:
        <dbooth> Present: dbooth jonathan mary
        <dbooth> Present+ amy

Agenda: http://lists.w3.org/Archives/Public/public-html-media/2013Jul/0008.html
Found Date: 09 Jul 2013
Guessing minutes URL: http://www.w3.org/2013/07/09-html-media-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]