Re: [rtcweb] Another reason not to use SDP (was: Draft agenda for IETF 87)

[Matt Fredrickson <[e-mail]>]

On Fri, Jul 12, 2013 at 3:23 PM, Roman Shpount <[e-mail]> wrote:

> On Fri, Jul 12, 2013 at 2:28 PM, Martin Thomson <[e-mail]>wrote:
>
>> That depends entirely on the nature of the instructions.  Allowing for
>> extensibility is always really, really hard, but maybe there is enough
>> experience with SDP now that this sort of scenario could be supported.
>>
>>
> Unless my past 10 years of experience building SIP systems mislead me, the
> current way of dealing with SDP extensibility is to define all features
> supported by network or device, go through the interop with every new
> network or device you need to connect to, and then, when in doubt, put an
> SBC in front of it. We currently need to go through interop testing every
> time we setup a new connection with a new VoIP service provider, and this
> is for simple PSTN to VoIP G.711+RFC2833 calls. Adding an unusual codec,
> like getting a direct interconnect in AMR-WB to cell phone networks, or
> SILK to Skype, requires years of testing and still not typically available
> commercially. SIP interworking for anything more complex (like even simple
> video end points) is virtually non existent. So, I guess, we are building
> on a very solid foundation of interprable solutions which require extensive
> testing after every sneeze.
>

I think much of it goes back to supporting a specific subset of SDP for a
core set of use cases.  In the Asterisk world, we generally try to support
nearly anything and everything that operates within our set of core use
cases.  But once you trespass too far outside of them, you can only fall
back to support the parts of the session description that are interpretable
by your implementation.

Given all of the different devices that we have come to interoperate with
over the years (just with SDP, not even necessarily with SIP or other
higher layer signalling), I too fear that a lack of specific restraints on
what is permissible within SDP as utilized within WebRTC will cause
interoperability and fragmentation problems, at least in more advanced use
cases.

Matthew Fredrickson