{minutes} HTML WG media telcon 2012-07-10 - roll call, previous meeting minutes, baseline documents, object oriented api proposals, new bugs filed, action items

HTML Media Task Force Teleconference

10 Jul 2012



The minutes are available in HTML here: http://www.w3.org/2012/07/10-html-media-minutes.html

or in plain text, below.



------------------------------------------------------------------------------

Agenda http://lists.w3.org/Archives/Public/public-html-media/2012Jul/0015.html

See also: IRC log http://www.w3.org/2012/07/10-html-media-irc

Attendees

Present

suzie, duncanr, +1.858.880.aaaa, johnsim-microsoft, pladd, acolwell, +1.720.934.aabb, BobLund, markw, ddorwin, adrianba, paulc, yang, [Microsoft], glenn, johnsim, BillyWatts, Tom_Handal, Clarke

Regrets

Chair

Paul Cotton

Scribe

johnsim



Contents

Topics

1.  role call

2.  previous meeting minutes

3.  review of action items

4.  base line documents

5.  object oriented api proposals

6.  new bugs filed



Summary of Action Items

<trackbot> Date: 10 July 2012

i can scribe

if you provide me the correct instructions

thanks

i am changing phones

<paulc> this conference is media

zakim who am i

zakim microsoft.a is me

<glenn> scribenick: johnsim

<Clarke> +Clarke_Stevens

<glenn> 720 is colorado





-------------------------

1. role call

scribenick johnsim-ms





-------------------------

2. previous meeting minutes

<paulc> http://www.w3.org/2012/06/26-html-media-minutes.html

<paulc> Noted



-------------------------

3. review of action items



-------------------------

4. base line documents

<paulc> Noted



-------------------------

5. object oriented api proposals

two bugs 16612, 16613

<paulc> https://www.w3.org/Bugs/Public/show_bug.cgi?id=16612

<paulc> https://www.w3.org/Bugs/Public/show_bug.cgi?id=16613

<ddorwin> have to dial back in apparently

<paulc> IRC FAQ: http://www.w3.org/2001/12/zakim-irc-bot.html

<glenn> http://www.w3.org/2002/03/RRSAgent

<glenn> http://www.w3.org/2005/06/tracker/irc

<adrianba> Agenda: http://lists.w3.org/Archives/Public/public-html-media/2012Jul/0015.html

<paulc> API design proposal #1 http://lists.w3.org/Archives/Public/public-html-media/2012Jun/0136.html

ddorwin: 136 was based on conversation two weeks ago

... key request generate session object on media element

... didn't address reuse

... reasons for separateing key creation from media elements

<ddorwin> separating key creation: http://lists.w3.org/Archives/Public/public-html-media/2012Jun/0134.html

ddorwin: with a constructor, not on lifetime of underlying media element, and may help with key release, etc.

<paulc> API design proposal #2 (using a MediaKeySession constructor) http://lists.w3.org/Archives/Public/public-html-media/2012Jun/0143.html

ddorwin: proposal #2, modification of original proposal, to use a constructor, that seemed to make sense,

... 2b supercedes because 2 would be too hard to implement

<adrianba> #2b -> http://lists.w3.org/Archives/Public/public-html-media/2012Jul/0002.html

<ddorwin> 2b: http://lists.w3.org/Archives/Public/public-html-media/2012Jul/0002.html

eric: 2b, post response in email

<paulc> Yang's questions are in: http://lists.w3.org/Archives/Public/public-html-media/2012Jul/0031.html

eric: describe for 2b, second proposal makes sense from api point of view. Question: why not associated with media element - add a session -

<yang> Why can not, can we use some method like addSession to add a session to a media element?

<ddorwin> addSession would not provide "strong association"

<yang> The key lookup will happen in CDM, so if we bind a session with a media element, and add key to the session, the key is passed to CDM, there will be no difficulty if only 1 key is needed for decryption, if for multiple keys for the same media element, I think the CDM only need 1 key at the same time to decrypt the media content, we may keep the principle, that 1 session for 1 key, and each time key change, the session removed and recreate, so there will be

ddorwin: principle in last meeting, one init data is one session

... when you go to decrypt a frame, you don't have session id

so you need to look at all sessions to find the key id

sessions mean nothing at time you do a decryption

only for key exchange

yang: session has multiple keys

ddorwin: media element can have multiple sessions

... multiple streams, audio and video using different keys, and you can have key rotations, perhaps on different sessions

yang: different tracks - different keys

paulc: not only possibility, not obvious what the keys are meant for, not pre-determined

ddorwin: CDM would be told what KID to use for a given block and look up in some table and why 2B was for a single table for a media element if all sessions you wanted to use were associated in some way

... proposal 1 had a strong association, but there are limitations, 2 had advantages

adrianba: a couple of things in the proposal, perhaps syntax details, aren't quite clear to me.

adrianba; media element has media keys object, but then you have the media key interface with a constructor which passes in the key system string

and it is not clear how that constructor gets called in the call example, not a (new) instance - key system string a bit like an index

ddorwin: the example has a mistake should say new media keys

... not intending to index anything

... unless to create from an element there has to be some sort of assigning.

... get at directions, debate specifics, high level comments, preference for 1, 2, 2b

<markw_> 2b looks good to me

adrianba: in broad terms, we ... just like we agreed on media source spec the object oriented approach is something we want to do, we should update the spec with this proposal, the 2b proposal is the best one to go with right now

<yang> 2 only for key session without collection right?

adrianba: more changes, but easier to review when it is in the document

<BillyWatts> agreed... 2b seems like a good direction

ddorwin: +1 on 2b

paulc: we now have 4

... may take a while to settle it down, but anyone object to starting from 2b ?

<yang> +1 on 2b, but need detailed

paulc: time to edit?

ddorwin: some major rewrites of algorithms, should be able to do it in the next two weeks

paulc: best use of our time now is to look at other bugs in the agenda or some other mechanism? ideal to be done in one week to give people chance to review before next call

ddorwin: see where we are at the next weeks call

ddorwin and the editors have offered to revise based upon proposal 2b

<paulc> #2b -> http://lists.w3.org/Archives/Public/public-html-media/2012Jul/0002.html

<paulc> ACTION: ddorwin and editors to implement proposal 2b by revising EME specification [recorded in http://www.w3.org/2012/07/10-html-media-minutes.html#action01]

<trackbot> Created ACTION-1 - And editors to implement proposal 2b by revising EME specification [on David Dorwin - due 2012-07-17].

at the media source meeting next week spend a minute to see where we are with the update to EME



-------------------------

6. new bugs filed

paulc: in order of agenda

<paulc> [Bug 17615] New: Support for different CDM communication models https://www.w3.org/Bugs/Public/show_bug.cgi?id=17615

adrianba: request in this bug is to support the model where the decoding is part of the media element and we actually have a separate

<paulc> Separate bug: https://www.w3.org/Bugs/Public/show_bug.cgi?id=16543

adrianba: bug, 16543, based on feedback from david singer, who pointed out that it is already possible with html5 to take care of content protection without any additional spec

... make changes to introduction for programmatic support for interacting with the content protection module and make it explicit that you can already do this in the media element

... proposed resolution, is to close it, as 'works for me'

<paulc> Proposed status is WORKSFORME with no changes

Yang; for this bug, two different CDM models, i noticed, UA and the CDM, it is not a javascript api

<adrianba> +1 - this is an implementation detail in the browser

<yang> and the retrieveKey does not need to be defined in W3C, it is not a JS API.?

paulc: editors should take action on this bug

... next 17658

<paulc> [Bug 17658] New: need procedure for selection of Key System https://www.w3.org/Bugs/Public/show_bug.cgi?id=17658

Yang: no clear discripion of how UA detects Key System

<yang> f media element's source attribute has a keySystem attribute then we use this as selected keySystem, this is OK. If no selected KeySystem, the specification says "jump to Key Presence" The question is : At "jump to Key Presence" is a else if of "selected key system", so here we have no idea about key system, but in "jump to Key Presence", we are handling key already exist, so why before key system selected, we can have key? We may need let a

ddorwin: mark and i replied. anything still unclear after the email and bug reply?

<paulc> See replies including: https://www.w3.org/Bugs/Public/show_bug.cgi?id=17658#c3

ddorwin: important part, UA never selects the key system, application selects

... with common encryption , you may need key system specific data in the file, but that is a separate issue

paulc: Yang, are you satisfied by the answer - UA does not pick the key system?

... if something missing in the algorithm, should suggest text.

yang: discuss offline in email

<scribe> ACTION: continue bugzilla and email discussion of 17658 [recorded in http://www.w3.org/2012/07/10-html-media-minutes.html#action02]

<trackbot> Sorry, couldn't find user - continue

paulc: next 17660

<paulc> [Bug 17660] New: need token relative with user identity for a new generateKeyRequest parameter https://www.w3.org/Bugs/Public/show_bug.cgi?id=17660

paulc: comments?

ddorwin: user identify in generate key request, perhaps part of license request, the thinking in the current proposal combine with init data

... so this is handled by the current specification and trying not to add any other features but open to discussion

yang: token outside of init data, perhaps downloaded from web server, token can passed to the CDM with the init data and other things

ddorwin: the application can combine token with init data in the current APIs, so looking for comment if explicit parameter like you propose is something people are interested in

adrianba: when we talked about this topic before original proposal, lots of different options for various data we might want to supply here, user token, or something in the applicaiton

andrianba: and ended up saying init data is a large block of data, a binary blob of data, we will need some more implementation experience to determine if that is the most appropriate whaay of doing this

adrianba: need to make more progress and experience/implementation before proposing that. so recommend you can combine in an application-specific way into init data and if we decide

... that is too complicated we can change the api

paulc: how to dispose of this bug. held for now? until we have more implementation experience?

adrianba: we could resolve it like that.

yang: okay, we can hold for now

paulc: Editors to resolve as "will not fix" for now, it will be easy to re-raise if experience indicates a single binary blob is not meeting the use cases



Summary of Action Items

[NEW] ACTION: continue bugzilla and email discussion of 17658 [recorded in http://www.w3.org/2012/07/10-html-media-minutes.html#action02]

[NEW] ACTION: ddorwin and editors to implement proposal 2b by revising EME specification [recorded in http://www.w3.org/2012/07/10-html-media-minutes.html#action01]



[End of minutes]



--------------------------------------------

John C. Simmons | Media Platform Architect | Microsoft Corporation | direct 425-707-2911  | mobile 425-269-5759

Received on Tuesday, 10 July 2012 18:27:17 UTC