Re: [rtcweb] SDP Security Descriptions (RFC 4568) and RTCWeb

[Dan Wing <[e-mail]>]

On Apr 30, 2013, at 7:51 AM, Iñaki Baz Castillo <[e-mail]> wrote:

> 2013/4/30 Dan Wing <[e-mail]>:
>>> - DTLS-EKT-SRTP:  This requires a gateway sending like "re-INVITE" for
>>> common operations as multimedia session transfer, which involves the
>>> gateway becoming both a media gateway and a complex signaling B2BUA
>>> (and we hate that, right?).
>> 
>> That re-INVITE is necessary because Security Descriptions changed the SRTP key, and Security Descriptions needs to send that re-INVITE.  It is not the fault of DTLS-SRTP-EKT, because it doesn't send a re-INVITE when the SRTP key changes (EKT handles key changes).
>> 
>> If the objection is the re-INVITE when SRTP keys change, the solution is eliminating Security Descriptions.
> 
> 
> The "solution" cannot be a media gateway that also must behave as a
> complex signaling B2BUA originating re-INVITEs by itself, since that
> means that a simple signaling pure proxy scenario is unfeasible.

Webrtc has already made decisions that make a pure proxy unfeasible with most legacy SIP networks:  ICE-Lite, RTP/RTCP multiplexing on the same port, and bundling. Those are trade-offs against interoperability but in favor of a better webrtc design.
 
I would rather the complexity and security weakness of Security Descriptions be confined to the interworking device, rather than exist forever in webrtc.

-d


> 
> 
> 
> --
> Iñaki Baz Castillo
> <[e-mail]>