Re: [dispatch] Proposal for a new WG: Privacy Enhanced RTP Conferencing (PERC)

[Christian Groves <[e-mail]>]

Hello Paul and Adam,

On 17/04/2015 4:59 AM, Paul Kyzivat wrote:
> On 4/16/15 2:29 PM, Adam Roach wrote:
>> On 4/16/15 13:23, Paul Kyzivat wrote:
>>> The advertisements are send over a data channel.
>>
>> Is this like a "data channel" in the same sense that WebRTC uses that
>> term?
>
> Yes. Chosen in order to facilitate the possibility of WebRTC clients 
> for CLUE.
>
>> If so, I think there are some good ways to handle this in line
>> with PERC's goals (there are reasonable ways to ensure that the data
>> channels terminate at the trusted key management node rather than the
>> mixer).
[CNG] The CLUE datachannel currently is terminated at the mixer. It 
needs to have access to the Advertisements in order for it to formulate 
the Advertisements that sent to other CLUE enabled endpoints.
>
> I fear we are talking past one another.
>
> Conceivably it would be possible to separate the signaling logic of 
> clue from the media mixing logic, so that there could be a *trusted* 
> entity that manages the keys and also does the mixing of clue 
> advertisements, but that offloads the actual media "mixing" (switching 
> in this case) to an untrusted media box.

> But it will require some careful thought, and probably changes to clue.
[CNG] It depends on your ambition level. CLUE v1 can still be used 
albeit in a simpler form. E.g. An endpoint using PERC probably shouldn't 
send XCARD information information about the participants. Basically it 
shouldn't send any metadata it doesn't want a mixer to access. There's 
still quite a few attributes that would be useful for capture selection 
and rendering that I couldn't see a problem with a mixer having access to.
If you want to pass XCARD information etc. securely through a mixer to a 
remote CLUE endpoint via CLUE then that's going to need work in CLUE.
>
> I guess that would be a CLUEv2, and we shouldn't get too hung up on it 
> right now.
[CNG] I do think its still worthwhile discussing as part of the 
architecture how mixing is controlled in a PERC architecture. I think 
its important we're on the same page. CLUE may have a role here.

BTW I do support this work.

Regards, Christian

>
>     Thanks,
>     Paul
>
> _______________________________________________
> dispatch mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/dispatch
>