HTML Media Task Force Teleconference -- 07 Oct 2014

<trackbot> Date: 07 October 2014

F2F information for Oct 30-31: https://www.w3.org/wiki/HTML/wg/2014-10-Agenda

<bevans> +q

<scribe> Agenda: http://lists.w3.org/Archives/Public/public-html-media/2014Oct/0013.html

<scribe> scribenick: paulc

Roll call

HTML WG F2F meeting plans, Fri Oct 31

See http://lists.w3.org/Archives/Public/public-html-media/2014Oct/0007.html for Paul's plan to meet Fri AM Oct 31

Bug 26776 - Diagnosing and resolving CDM errors needs a numeric systemCode (deleted with MediaKeyError)

https://www.w3.org/Bugs/Public/show_bug.cgi?id=26776#c8

MediaCapture have the same problem as we do.

Can we get a volunteer to work with them on this topic.

Their proposal is "I suggest we combine forces and ask for a (long or DOMString) secondaryArgument on DOMException at least as a stopgap solution "

Jerry will take on this coordination and will use the IE participants on MediaCapture to make the connection.

Mark: This would work for us but not sure if the DOM people will like this.
... We need to engage people like Anne to see if there is any chance this would be acceptable.

Bug 25966 - Use ArrayBufferView and ArrayBuffer instead of Uint8Array in APIs

https://www.w3.org/Bugs/Public/show_bug.cgi?id=25966#c4

David: I reopened this and it will be fixed after the Respec editing is done.

Mark: There is an IDL facility for this.

David: That is why I re-opened this bug.

<joesteele> +q

Bug 25923 - isTypeSupported should be asynchronous

https://www.w3.org/Bugs/Public/show_bug.cgi?id=25923

Discussion thread: http://lists.w3.org/Archives/Public/public-html-media/2014Oct/0002.html (modified proposal)

David; This solves the original Mozilla bug and is okay with other developers.

scribe: same axis as used by Web MIDI
... also solving other issue on capabilities not supported by a list of strings
... you provide a list of items supported and you are returned the first one supported
... still some outstanding questions
... questions will become clearer as spec text is written

David: I am looking for feedback but the proposal solves not only the original bug but other problems
... will go in this direction without more feedback
... will be implemented before the F2F but after the Respec editorial work

Joe: Will there a list of video capabilities?

David: Same as in isTypeSupported.

Joe: Not constrained and that is what I wanted to know.

<joesteele> thanks

Jerry: Why did you split into audio and video systems?

David: Genernally wanted to permit different protections. Proposal referenced a former bug.
... May be someone wants greater protection for 24-bit audio.
... Did this for flexability

<joesteele> examples would be Dolby, DTS -- both are concerned more about audio protection

Bug 26838 - Normatively address vulnerabilities related to initData contained in media data

skipping this item for now

Bug 26372 - Report issues/events not related to a specific method call

https://www.w3.org/Bugs/Public/show_bug.cgi?id=26372

Discussion thread: http://lists.w3.org/Archives/Public/public-html-media/2014Sep/0023.htm

David: Trying to figure out how to report all errors
... eg keys not be usable or not available

<joesteele> http://lists.w3.org/Archives/Public/public-html-media/2014Sep/0023.html

David: how do we want to provide the status
... how do we do this in ECMAScript
... Current direction seems to be map-like as in new WebIDL
... or object.observe
... probably go with map-like thing which you can iterate over
... Not much discussion but the map-like proposal looks good

<markw> +1 to maplike and event giving the key ids added or changed

Joe: my objection to the map-like direction is how does a developer observe changes to the map

David: There would be a simple event like "keyschange"

Joe: Going down map direction. The event will point us to the map change?

David: Observable would give you a list of changes but you would have to figure out your own state.

David; Option 4 was an alternative - report all the changes ie poor man observable

David: Basic facility would be "something has changed" and the app needs to figure this out

Mark: It would be nice if the keys changed event gave an exact list of changed keys

Joe: I may object to the map facility since it creates data structure that might not be useful ie for key transitions
... Not against the map but it may be overhead that we don't need

David: The IDL has lots of facilities to use the CDM provided information
... Modelled after ECMAScript maps which developers should be used to

Joe: Looks like it has all the accessors needed. Not sure how easy it is to construct.

paul: what is timetable for this one.

David: After 26372 and MAYBE before the F2F.

<joesteele> I will defer to the browser implementors on this one -- not something I think would be always used, but if overhead is low I have no problem with it

Bug 26887 - Allowing license servers and CDMs to control data persistence and secure release

https://www.w3.org/Bugs/Public/show_bug.cgi?id=26887

paul: Most recent discussion has been in the bug in the last 2 weeks.

Jerry: I asked for this to be on today's agenda.
... I raised the concern that finding a single solution for both problems might be hard
... There is sitll disagreemen on whether the proposed model belongs in EME
... There are lower level problem outstanding that can be fine tuned BUT
... we need to get agreement on the higher level model

<ddorwin> Nobody has argued that the license can't specify whether the license is stored. The current spec just requires that a "persistent" session be used.

Mark: Not correctto think this as two competing models
... the spec does not preclude the license containing extra information
... current status is that the app and license need to specify if the license can be stored
... other distinct issues are a) can the CDM re-use the stored information

<ddorwin> +1 to what Mark said

Mark: I don't have a general problem with re-using InitData but there are use cases where the app needs to clearly state how a license should be reused.

Jerry: Maybe an attribute to force a new license to be acquired

Joe: I agree with Mark. I would need or would like to see a specific flag
... the default case we would use cached licenses so there needs to be a flag to go around that
... we don't have those options today
... load is part of the that flag but not the whole story
... I will add that into the bug (been busy with other items)
... For us LoadSession and CreateSession cause us to cache the InitData
... This bug tries to split those apart.

Jerry: The proposal is to eliminate LoadSession
... we concentrate on CreateSession and let it re-use an existing license and skip some steps

<ddorwin> LoadSession is important for multiple use case. It doesn’t have to be removed to solve your scenario.

David: I have already identified issue with there being no sessions

<ddorwin> When thinking about solutions: Could have a SessionId lookup based on initData. Gives the application control with the alternative lookup you want.

paulc: self-scribing like David is doing is a great model for others to use

<ddorwin> Nobody has argued that the license can't specify whether the license is stored. The current spec just requires that a "persistent" session be used.

<markw_> There are cases where there are multiple licenses cached for the same content item. The application needs to be able to specify which one.

<ddorwin> +1 to markw_. A unique ID is the only way to enable this.

Mark: Session id was the way to specify the license to use
... you could have different kinds of licenses
... we need to be able to handle these use cases

Jerry: What is the use case for different license and how does it matter?

Mark: Maybe there is a preview license with restrictions and a full license that has side-effects when used.
... you need to be careful what license to use

Joe: I guess the issue I have with LoadSession the way it is, is can the Session loaded go off and get new licenses?
... How do I tell the CDM to NOT use existing licenses?
... Another point is that you need to be able to stop the CDM using cached licenses.

David: Let's look at the existing rules in the document. If there is no Session then it is not possible to use cached licenses.

Joe: The exact issue is "what is the spirit in the spec?"

<markw_> I don't think there is a disagreement about whether there are such things as cached licenses. But you need to have a session to use them.

Joe: The spec permits the CDM to cache licenses ...

David: I disagree that the spec permits that.
... if we are missing that normative text then we need to fix that

Mark: Are you saying that the CreateSession permits you to use a cached license.

Joe: yes

David: We seem to be disagreeing whether you can use a cached license or have to specifically load it.

Jerry: the model Joe is asking is the one that this bug is asking for

<joesteele> my problem is that generateRequest was added as a solution to this without fully answering everyones issues with it

<joesteele> generateRequest was not acceptable as a solution to this issue

Jerry: I got some feedback today

<markw_> generateRequest (aka createSession) presently doesn't allow re-use of cached data because it requires a licenserequest message to be generated

Jerry: MSFT participants will try to modify the proposal and ensure wide discussion

<markw_> for me it could be fine to allow createSession (rename init?) to re-use cached data, but we have to solve the problem of disambiguating multiple stored records keyed of the same initData and the problem of allowing the app to prevent re-use

Bug 26811 - Separate definitions of Initialization Data Types from Stream Format parsing

Bug 26811 is blocking Bob Lund's possible work on Bug 26738

paul: When will 26811 be done?

David: It is still a lower priority that the bugs that could shape the API.

Paul: 26811 will probably not be done before the end of Oct

David: You can probably get some of the content ready

Bob: I would prefer to wait so that I don't do double work.

- DRAFT -

HTML Media Task Force Teleconference

07 Oct 2014

Attendees

Contents