Re: [MMUSIC] SCTP-SDP: Virtual Connection impact

[Roman Shpount <[e-mail]>]

On Tue, Mar 31, 2015 at 1:00 PM, Christer Holmberg <
[e-mail]> wrote:

>  >The note tells what does not trigger setup of the new DTLS connection
> in case of ICE. It does not specify what does. I also >think it
> is ambiguous. Current language implies that all candidate pairs associated
> with SCTP association are the same >transport.  Does this mean that if SCTP
> connection is not re-established and ICE is used, new DTLS session is not
> started? >Or does it mean that a new DTLS connection is setup for each new
> ICE session? What about ICE restart (i.e. ufrag change >between the same
> end points when the same candidate sets are reused on both sides)?
>
>
>
> The idea is that a change of the fingerprint would trigger a new DTLS
> connection.
>
>
>
> But, in order to avoid the receiving-packets-associated-with-old-keys
> issue (raised by yourself earlier) the suggestion was that the transport
> 5-tuple also needs to change.
>
>
>
> I agree some wording is needed about that.
>
>
>
The idea was also that fingerprint can only change if underlying transport
changes as well. I think we need to define what this means in case of ICE,
since 5-tuple change, when ICE is used, is not necessarily a transport
change (as in cases when it is a change between the candidate pairs). I
think the intent is, that in case of ICE, DTLS fingerprint can change only
if new ufrag is used (new ICE session) and entire candidate list changes
(not an ICE restart on existing candidates). I would really like to see if
the rest of the people on the list agree and this decision is documented
somewhere.
_____________
Roman Shpount