IETF
avtext@jabber.ietf.org
Thursday, November 5, 2015< ^ >
Jonathan Lennox has set the subject to: AVTExt working group
Room Configuration
Room Occupants

GMT+0
[03:40:29] Meetecho joins the room
[03:56:30] Brian Rosen joins the room
[03:57:37] Bernard Aboba joins the room
[03:59:01] <Brian Rosen> I will be the jabber scribe for the meeting.  If you want me to address the room, please preface with mic:
[04:01:17] hta joins the room
[04:06:51] Randell Jesup joins the room
[04:07:17] Maire Reavy joins the room
[04:07:29] Mark Harris joins the room
[04:08:53] Suhas Nandakumar joins the room
[04:11:03] <Bernard Aboba> mic: Have you considered addition of TL0PICIDX, which is supported in H.264, VP8 and VP9?
[04:16:12] Suhas Nandakumar leaves the room
[04:16:27] ben joins the room
[04:16:47] Suhas Nandakumar joins the room
[04:16:56] <Brian Rosen> okay, how do you pronounce TLOPICIDX?
[04:18:56] Suhas Nandakumar leaves the room
[04:19:09] Suhas Nandakumar joins the room
[04:19:12] <Randell Jesup> TL0PICIDX, of course ;-)
[04:26:24] iwashi joins the room
[04:29:22] <Brian Rosen> Tee Ell Zero Pick Index, thanks
[04:31:23] <Bernard Aboba> mic: RID is for use with simulcast, but not SVC, no??
[04:31:39] <Bernard Aboba> If so, it would be orthogonal.
[04:33:00] lavanya HM joins the room
[04:34:10] Suhas Nandakumar leaves the room
[04:34:25] <Brian Rosen> woops, missed it, sorry
[04:36:38] lavanya HM leaves the room
[04:36:40] mala solanki joins the room
[04:38:09] shashank kaushk joins the room
[04:38:18] shashank kaushk leaves the room
[04:40:38] swati chawla joins the room
[04:44:54] ronit sharma joins the room
[04:46:01] tanya kaur joins the room
[04:46:39] nidhi jian joins the room
[04:48:26] sangeeta berry joins the room
[04:48:39] salman khan joins the room
[04:49:09] anupam singh joins the room
[04:49:41] shahrukh khan joins the room
[04:49:45] shahrukh khan leaves the room
[04:51:54] amit pamnani joins the room
[04:52:37] Malay Bandyapadhyay joins the room
[04:52:40] sonam kapoor joins the room
[04:52:42] anil sharma joins the room
[04:52:54] amir khan joins the room
[04:54:10] Malay Bandyapadhyay leaves the room
[04:56:31] anjani bhatnagar joins the room
[04:56:35] rani mukherjee joins the room
[04:57:00] arjoon kapoor joins the room
[04:57:05] abhinav diwan joins the room
[04:57:43] <rani mukherjee> yay
[04:58:15] <rani mukherjee> wassup shahrukh , amir, salman... howz friendship going?
[04:58:21] <rani mukherjee> hummmmm
[04:59:33] prateek deka joins the room
[05:00:19] sunil chandolia joins the room
[05:00:49] prem ratan joins the room
[05:01:39] adolf hitleer joins the room
[05:01:46] Anish Peter joins the room
[05:02:58] Anish Peter leaves the room
[05:03:13] <Bernard Aboba> mic: Agree with Mo - reordering will go against you here.
[05:03:21] mala solanki leaves the room
[05:03:35] <Bernard Aboba> Just because the receiver gets 3, doesn't mean that 2 didn't arrive - it won't know that until later.
[05:11:50] <Randell Jesup> mic: Comment to Colin's (first) comment: This allows you to implement a class of congestion alogorithms by only modifying one end (in this case, the sender's end.) You could invert this, and run the CC at the receiver, but you still need a report-back RTCP value, and send-time RTP extensions.  The downside is you wouldn't know the size of any lost packets.  To Colin's recent comment, the more bits of data to feed into the delay estimator, the faster it can converge and detect increasing/decreasing delay.  I agree with Michael, I tihnk. Perhaps a feedback-per-stream, with the sender taking all these items and merging them into a single estimate.  I'm not sure that the feedback message itself has to be across all.   Sorry for the long message...  and I may have lost context while writing it it and listening
[05:20:52] <Randell Jesup> mic: One big advantage to receiver-side is that it *only* needs to know the relative send times, and (less critically) if there's a gap (we'll find that out via existing seq numbers).  The send side really only needs to know what rate it has to divide up.  If you want to merge the congestion controllers for the streams, you can just use the send-time header extension (and save bits on each packet).  Then send the sender a small RTCP with the result of the delay correlation and loss.   And this keeps normal RTP semantics...  Or you can do as Colin says, if you want to run it all at the sender side.
[05:23:13] <Randell Jesup> mic: this is like what Michael says - with a lot less data having to be sent back to the sender, I think.
[05:23:53] <Brian Rosen> 5th on line for mic
[05:23:58] <Randell Jesup> thanks
[05:24:14] <Brian Rosen> np
[05:28:57] jesup joins the room
[05:34:21] ben leaves the room
[05:35:38] Bernard Aboba leaves the room
[05:36:06] Bernard Aboba joins the room
[05:41:41] <Bernard Aboba> It seems reasonable.  How realistic it is is another story ;)
[05:52:15] ben joins the room
[05:54:31] ben leaves the room
[05:55:27] Maire Reavy leaves the room
[05:55:28] iwashi leaves the room
[05:55:50] Meetecho leaves the room
[05:56:55] hta leaves the room
[05:57:08] Brian Rosen leaves the room
[06:00:09] anjani bhatnagar leaves the room
[06:05:07] Brian Rosen joins the room
[06:06:38] Brian Rosen leaves the room
[06:06:38] Brian Rosen joins the room
[06:09:21] Randell Jesup leaves the room
[06:09:21] Bernard Aboba leaves the room
[06:09:21] anupam singh leaves the room
[06:17:54] ben joins the room
[06:22:03] ben leaves the room
[06:29:12] hta joins the room
[06:35:53] <jesup> hum understand
[06:36:45] jesup leaves the room
[07:10:16] hta leaves the room
[07:15:08] Brian Rosen leaves the room
[07:16:05] Brian Rosen joins the room
[08:10:18] Brian Rosen leaves the room
[08:20:14] hta joins the room
[09:02:23] hta leaves the room
[09:07:45] hta joins the room
[09:31:24] hta leaves the room
[11:19:27] hta joins the room
[14:02:01] hta leaves the room
[21:48:50] hta joins the room
[22:15:38] hta leaves the room
[22:30:25] hta joins the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!