RE: [AVT] Comments on draft-ietf-avt-mime-h224-02.txt

["Gunnar Hellstrom" <[e-mail]>]

Roni and Kae,

Your discussion on transport of FECC takes analog turns as the
discussion on real time text does.
Real time text is defined for RTP transport in RFC 4103. It has its MIME
type text/t140.
In a video call there may be long moments when the participants just use
video and audio, and thus there is no traffic in the text stream. But
when someone types it should be transferred in real time, and it can
occasionally be substantial amounts of text being transmitted.

It is sometimes proposed that we should multiplex text in the audio
channel. In the early beginning of the discussions it was also proposed
to send it in the call control channel.

The result has always been that it is clean and according to SIP
architecture to have the text as a separate RTP stream. In reality there
are the risks and problems you mention with FW/NAT traversal. Not much
though. Proper SIP aware NAT routers tend to keep the path open without
a need for keep-alive traffic.

I would like to see the NAT/FW traversal discussions take this kind of
intermittent media in account and not close paths on timeout because of
lack of traffic during sessions.
And I agree with Roni that intermittent use of connected RTP sessions is
a proper way of handling this kind of real time need.

Regards
Gunnar
-------------------------------------------
Gunnar Hellstrom
Omnitor AB
Renathvagen 2
SE 121 37 Johanneshov
SWEDEN
[phone]
Mob: [phone]
www.omnitor.se
[e-mail]
--------------------------------------------


>-----Original Message-----
>From: [e-mail] [mailto:[e-mail]]On Behalf Of
>Even, Roni
>Sent: Wednesday, August 03, 2005 8:58 AM
>To: kae
>Cc: IETF AVT WG
>Subject: RE: [AVT] Comments on draft-ietf-avt-mime-h224-02.txt
>
>
>Kae,
>Thanks for the comments, I want you to note that the h224 mime type is
>defined for having interoperability with current video conferencing
>terminals through a gateway. It does not define a payload. This will
>enable a signaling gateway to work since the FECC will flow end to end.
>The payload is described in H.323 annex Q and not in this
>specification.
>
>We intend to look for other ways for FECC and are open to suggestions.
>As for your questions see online
>Roni
>
>-----Original Message-----
>From: [e-mail] [mailto:[e-mail]] On Behalf Of
>kae
>Sent: Wednesday, August 03, 2005 6:17 AM
>To: Even, Roni
>Cc: 'IETF AVT WG'
>Subject: [AVT] Comments on draft-ietf-avt-mime-h224-02.txt
>
>Roni,
>
>I am not sure if we should continue using Annex Q for FECC
>particular in
>SIP based systems. My concerns are:
>
>1. Adding a new RTP stream will make NAT/FW traversal more difficult
>than it is today even with helps from ICE/STUN/TURN.
>
>I agree that every RTP stream needs to traverse NAT/FW, so are you
>suggesting we will not use more than one media in a call?
>BTW even if we open a different control channel for FECC (Using FECC in
>the signaling path, see your comment 4) we will have a channel signaled
>by SDP that will need to traverse FW/NAT
>
>
>2. From security point of view, we do not want to open too many ports.
>In additions, FECC is general used a couple times during a video call.
>Most of time there are no packet flows. Most of NAT/FW will close the
>channel after
>1 min of inactivity.
>
>The FECC channel can be opened only when there is a payload to send by
>using re-invites or kept open by having keep alive. This is general
>procedure and will work if the FECC is in a separate control channel or
>as H.323 annex Q payload. As for number of open ports, this is
>a general
>problem. There was a suggestion to multiplex streams by SSRC which did
>not progress. In a multimedia conference you may have 1 audio, 2 video
>and a data connection for each call not including the signaling.
>
>3. I think we should keep the principle that RTP is for medium not for
>control. If we allow non real time traffic to use RTP, it will make QoS
>more difficult. A lot of devices are marking DiffServ Code Point based
>on RTP. If we start to use RTP for non-real time stuff, QoS
>devices will
>have to look into payloads which is not scalable. (Please note that we
>cannot trust end point's DiffServ marking)
>
>The contribution is not defining a payload type but just a
>MIME type for
>H224. BTW FECC is real-time traffic and that is why it should go end to
>end. The feedback is based what you see in the received video so
>real-time is very important.
>
>
>4. Using Annex Q/H.224, it makes certain architecture assumptions. I
>think the medium processing function and call control/feature functions
>should be separate either logically or physically. Given the fact that
>control paths do not use RTP. This draft will break this principle
>(except the case controller and media processor are physically
>the same)
>and ask the medium process to process FECC functions. Do you think it
>will be easier just send a SIP based message to the controller
>directly?
>
>FECC is not call control, carrying it using SIP is wrong since we will
>use the real-time needed. You expect to see  the result of the movement
>fast. This is why you are trying to move the camera. The signaling path
>is not the way for it. Example of FECC usage that is common is for tele
>medicine where doctor watch an operation from remote and they control
>what they see. (You do not want the operating doctor to stop and move
>the camera).
>
>
>This is my humble opions.
>
>
>Kae
>
>
>_______________________________________________
>Audio/Video Transport Working Group
>[e-mail]
>https://www1.ietf.org/mailman/listinfo/avt
>
>_______________________________________________
>Audio/Video Transport Working Group
>[e-mail]
>https://www1.ietf.org/mailman/listinfo/avt



_______________________________________________
Audio/Video Transport Working Group
[e-mail]
https://www1.ietf.org/mailman/listinfo/avt