Re: [rtcweb] Layers in draft-jesup-rtcweb-data-00

[Justin Uberti <[e-mail]>]

I think we want one DTLS session for each transport. If we have independent
RTP sessions, we have N transports, and N DTLS sessions. If we multiplex
RTP/SCTP sessions, we have a single transport, and the same DTLS session is
used to protect SCTP traffic, as well as to get keys for SRTP traffic.

On Tue, Nov 1, 2011 at 11:19 AM, Michael Tüxen <
[e-mail]> wrote:

> On Nov 1, 2011, at 3:10 PM, Magnus Westerlund wrote:
>
> > On 2011-10-26 22:45, Michael Tüxen wrote:
> >>
> >> On Oct 26, 2011, at 10:28 PM, Hadriel Kaplan wrote:
> >
> >>> But rather this:
> >>>
> >>>       +------+                        +------+
> >>>       |WEBAPP|                        |WEBAPP|
> >>>       +------+------+------+          +------+------+------+
> >>>       | DTLS | Audio| Video|          | SCTP | Audio| Video|
> >>> +---------------------------+   +---------------------------+
> >>> | STUN | SCTP |S/RTP |S/RTP |   | STUN | DTLS |S/RTP |S/RTP |
> >>> +---------------------------+   +---------------------------+
> >>> |         Mux/Demux         |   |         Mux/Demux         |
> >>> +---------------------------+   +---------------------------+
> >>> |            UDP            |   |            UDP            |
> >>> +---------------------------+   +---------------------------+
> >>>
> >>> [Note: "S/RTP" = SRTP/SRTCP or RTP/RTCP, "Mux/Demux" = tiny logic to
> mux/demux]
> >>>
> >>> Because the audio/video streams may be using the same UDP port, right?
> >>>
> >>> And the two "S/RTP" boxes may be just one box depending on how the
> MMUSIC multiplexing decision turns out.
> >>>
> >>> So if we want to choose the left one, because we expect/want that
> someday the Operating System provides a SCTP/UDP stack in the kernel, and I
> think we do, could it do so while demuxing and letting STUN, RTP, and DTLS
> go up to the app layer?  (i.e., given a socket/BIO/FD model)  I have no
> idea about such things... just asking.
> >> This is not possible today. The demultiplexing seems to be specific to
> this scenario.
> >> Not sure it fits. For demuxing you use the first byte to distinguish
> STUN from DTLS and SRTP.
> >> The first byte us the high order byte of the source port. Once could
> require
> >> SCTP to use source ports with the high order byte > 192. That might
> work.
> >> However, you would need to get the Mux/Demux into the kernel. Could be
> done
> >> using a socket option. But I'm not sure it really fits. Maybe Randy has
> an
> >> opinion.
> >>
> >
> > Michael,
> >
> > I think one of the reasons there is discussion of use land
> > implementations of SCTP is so that you can do the above stack diagrams
> > with SCTP above UDP that is being shared for several purposes.
> >
> > I would also like to correct Hadriel's picture slightly:
> >
> >>>       +------+                        +------+
> >>>       |WEBAPP|                        |WEBAPP|
> >>>       +------+------+------+          +------+------+------+
> >>>       | DTLS |Audio & Video|          | SCTP |Audio & Video|
> >>> +---------------------------+   +---------------------------+
> >>> | STUN | SCTP | DTLS-SRTP   |   | STUN | DTLS | DTLS-SRTP   |
> >>> +---------------------------+   +---------------------------+
> >>> |         Mux/Demux         |   |         Mux/Demux         |
> >>> +---------------------------+   +---------------------------+
> >>> |            UDP            |   |            UDP            |
> >>> +---------------------------+   +---------------------------+
> >
> > I think the above indicating that the common RTP session, potentially
> > being a DTLS-SRTP keyed RTP session that can co-exist on the same UDP
> flow.
> >
> > To make the left one work, I think one has to have a source port where
> > the first byte value is 192-255 as the de-multiplexing table from
> > section 5.1.2 of RFC 5764 shows these to be the only available.
> >
> > Thus I think the options are:
> >
> > A. Left figure above. (source port must be in 49152-65535)
> >
> > B. Right figure above
> >
> > C. SCTP in its own UDP flow plus one UDP flow for each RTP session.
> >
> > If we would pick SCTP then I think we must have either A+C or B+C
> working.
> And you want one DTLS connection for key derivation of SRTP and a
> separate one for SCTP? Or must they be the same?
>
> Best regards
> Michael
> >
> > cheers
> >
> > Magnus Westerlund
> >
> > ----------------------------------------------------------------------
> > Multimedia Technologies, Ericsson Research EAB/TVM
> > ----------------------------------------------------------------------
> > Ericsson AB                | Phone  [phone]
> > Färögatan 6                | Mobile [phone]
> > SE-164 80 Stockholm, Sweden| mailto: [e-mail]
> > ----------------------------------------------------------------------
> >
> >
>
> _______________________________________________
> rtcweb mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/rtcweb
>