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

[Tim Panton <[e-mail]>]

On 29 Apr 2013, at 06:20, Binod wrote:

> I have been reading the discussion on this topic and I prefer
> supporting SDES as a keying method for WebRTC.
> 
> Not having SDES will have non trivial impact on interop. With
> EKT, there is a signalling complexity of sending re-INVITEs, which
> make the gateway complex. Without EKT, you need per-packet
> crypto  for media exchange, which is CPU intensive.

I've seen this asserted more than once, but I'd love to see a _current_ example where 
you really have an existing network of SRTP/ICE/BUNDLE/RTCP-MUX capable
legacy endpoints that you want to connect to webRTC without a media-level SBC or 
call recording. 

My fear is that people are just basing anti-DTLS opinions on the perceived difficulty of 
building such a network in the future. 

I'm ok with legacy interop as a secondary goal of this WG , but putative-future-legacy interop
is going too far IMHO, especially since it further complicates the already tricky problem of
defining interoperable SDP.

If it is a choice between adding complexity in a legacy gateway or every browser
I'd rather add it in the gateway.

T.