Re: [MMUSIC] SCTP AppID

[Salvatore Loreto <[e-mail]>]

On Mar 6, 2014, at 2:14 PM, Martin Thomson <[e-mail]>
 wrote:

> On 6 March 2014 13:36, Salvatore Loreto <[e-mail]> wrote:
>> I understood Martin was only proposing to look at how ALPN defines and handles the registry
>> and do something similar not using it. Am I wrong?
> 
> Yes, you are wrong.
I was sure to be wrong
using it would make my life easier and I would love to use it
 but I have still some concerns about it that maybe you can help me to solve

> 
> My intent was to say that you should be using the registry in an
> unadulterated form.
> 
> ALPN is being used now in a number of places to identify protocols.
can you point me to those?

> We've more or less unilaterally decided in httpbis that this means the
> entire protocol stack.
I know of the decision in httpbis
but as it is not written down clearly anywhere I was and I am not sure about 
leveraging on that decision.


>  I'm thinking that the protocol we're
> identifying here is the following stack of protocols:
> 
> Data channel protocol (the combination of the data chennel
> establishment protocol and the use of the text and binary PPIDs).
> SCTP
> DTLS
> UDP
> IP

another concern is about…
are we allowed to register also value that has DTLS and not TLS in the tower?

on more point is that WebRTC is a mess!!!
you can end up of a situation where you already have a DTLS association established 
because you need the key to crypto to be used in SRTP
and then at some point the user decide to  "reuse" the same DTLS association to run 
an SCTP association on top of it


> 
> Yes, there are cases where there are other users, or protocols perched
> on top of this tower.  Those could also be identified with an ALPN
> identifier too, if they wanted to.
> 
> One such higher-level protocol is the combination of this glorious
> protocol tower with SRTP.  That's another ALPN identifier.