At 09:11 AM 11/25/97 -0800, Toby Nixon wrote:
I think I've stated my position on the reflector regarding V.chat/T.chat -- that it is clearly a "data" application, and it ought to be a T.120 app and not ride directly on a logical channel or be in H.245 (at least in the context of H.323).
I tend to agree, except that so many people felt that T.120 is such a "burden" that it seemed that few would implment V.CHAT if it required T.120. That was the motiviation to allow it to be used without T.120, in a point-to-point mode.
I don't think I feel that way about camera control; it just seems so closely tied to the video media stream, and not really "data". It strikes me as more like chair/floor control than like chat.
When would you ever have multipoint camera control -- I mean, wanting to send camera commands to multiple endpoints simultaneously? I know you'd want to be able to separately control multiple cameras.
You'd use multipoint camera control anytime you are in a multipoint call and want to control the camera on a _particular_ far-end machine. This requires addressing of camera commands, and probably arbitration and locking if more than one person tries to control the same camera.
This could be done in H.245, but that would require the MC/MCU to both support the new H.245 messages (an upgrade) and do some message routing and arbitration to make it work. This is adding new functionality to the MC/MCU H.245 stack, when these functions are already in the T.120 infrastructure. So it seems simpler to just use T.120 when mutlipoint function is need. Since T.120 is considererd "a burden" by some, and isn't needed for point-to-point, we can just run the same protocol by itself in a H.245 LC (ala V.CHAT). Keeping the core protocol separate from H.245 allows the SAME protocol to be used both point-to-point (in a H.245 LC) and multipoint (in T.120). If we put it in H.245 directly, we'd end up with two sets of syntax to keep sychnronized in ITU-T, one in H.245 and another in T.120.
Anyway, I think we can break the camera control out of T.130 and make it a separate recommendation, and carry the PDU stream on a separate TCP connection negotiated using H.245. Basically, parallel what was done for H.224. Should it go through the MC, or directly to the controlled endpoint?
For point-to-point I'm not sure. For multipoint, it would go in T.120 to the T.120 peer (probably the MC).
In Sunriver, didn't we officially move T.130 from Q3 to Q14? That would seem to argue for presenting this to Q14 instead.
Would we want all of the remote device control stuff, including speaker volume, microphone gain and mute, etc.?
I would think so.
Do you want to move this discussion back out into public view and get input from others?
I'm cc:ing this to the H.323 list.
--Dave L.
-- Toby-----Original Message----- From: Dave Lindbergh [SMTP:lindbergh@itu.ch] Sent: Monday, November 24, 1997 4:26 PM To: Toby Nixon Cc: duckworthm@pictel.com Subject: RE: Inconsistent references to H.281
Toby,
At 02:37 PM 11/24/97 -0800, you wrote:
I'd be interested in working with you on this. Do we need some sort of proposal for a work item at the next SG 16 meeting? Or just bring in a
first
draft? I think this would go into Q14; do you agree?
You could bring in a draft, but it would probably be better to start with a general discussion in the group to see if there's consensus on this direction, before a lot of work goes into it. Maybe that will happen on the reflector.
I think we'd be interested in working with you on it.
We were thinking about putting in a contribution into the Q3 meeting in Geneva, since we thought it belonged in the Q3 group more than in Q14. You could argue either way, I've no strong preference.
What's the compelling reason for opening a separate logical channel for camera control instead of just issuing the commands directly in H.245?
The main reasons are (a) keep down the clutter in H.245 (already bad), and (b) make it easier to have a T.120 "wrapper" around the RDC stuff for multipoint use. We'd like to see a solution that doesn't require upgrading all the MCUs and gateways every time we add a new terminal feature; T.120 does this nicely for multipoint (there is no issue for point-to-point). This is the same path ITU took for V.CHAT/T.CHAT, it seems like a good model for this, too.
What do you think? Mark Duckworth here at PCTL has been looking at this from the Q3 viewpoint, I'm cc:ing him on this.
--Dave L.
-----Original Message----- From: Dave Lindbergh [SMTP:lindbergh@ITU.CH] Sent: Sunday, November 23, 1997 3:22 PM To: ITU-SG16@MAILBAG.INTEL.COM Subject: Re: Inconsistent references to H.281
Toby/all,
The H.224 codepoint in DataApplicationCapability.application.h224 of H.245 was put there for use with H.324; you can use H.224/H.281 to do far-end camera control in H.324. (H.224 itself negotiates the use of H.281; that's why you don't see H.281 in H.245.)
My feeling (and I think this has been the view of the group as a whole) is that H.224/H.281 won't work for H.323. The H.281 control model is timing based: You send a message to start moving the camera (presumably at some slow, fixed rate), then another to stop it. Even with H.320 or H.324, this requires the user to "lead" the camera to compensate for end-to-end round trip delay.
With H.323, any inconsistency in packet delivery times will make it impossible for the user to predictably control the camera. This could be partially fixed by adding timestamps to the messages, but only at the expense of buffering for the worst-case packet delivery time.
I think the RDC portion of T.130 is a much better approach: It is not timing based, but based on absolute positioning of the camera ("turn left 24 degrees").
Since T.130 is stalled, I suggest we think about extracting the RDC portion of T.130 (clearly needed for H.323) and approving it as a separate Recommendation (not necessarily using T.120). We could follow much the same approach as with V.CHAT/T.CHAT: A simple stand-alone protocol (the RDC ASN.1 structures from T.130) which can be used point-to-point in a H.245 LC, and a multipoint version that makes use of a T.120 "wrapper" (ala T.CHAT) around the same simple protocol.
--Dave Lindbergh
At 03:01 PM 11/21/97 -0800, you wrote:
We were poking around a bit for information on far-end camera control,
and
ended up searching for references to H.281 in most of the documents
being
decided in January. As a result, we uncovered some inconsistencies
that
may
require editorial correction of the decided texts of H.245, H.246, and H.323.
(1) H.245 section 2 refers to 1996 versions of H.281 and H.224. We can
find
only versions published in 1994. However, I note that although H.224
and
H.281 are listed as references, they are not actually referenced in
the
text
of H.245 anywhere, either by name or by the reference number. Either
these
references should be deleted from section 2, or the dates changed to "(1994)".
(2) H.246 refers to H.281 and H.224 in section A.9, but they are not included in the references in section 2. The solution here could be to simply remove section A.9, since it is "for further study"; if this is
not
done, H.281 and H.224 should be added to section 2.
(3) H.323 version 2 refers to H.281 and H.224 in section 6.2.7, but
neither
of these are included in the references in section 2. H.281 and H.224
should
be added to section 2. Note that H.323 version 1 had the same problem.
The reference in H.323 and the presence of DataApplicationCapability.application.h224 in H.245 seem to imply that H.281/H.224 can be used in H.323 terminals. It would seem fairly straight-forward to open a camera control data logical channel
associated
with a video channel and send H.281 commands. Does anybody actually do
that
today? Has it even been prototyped? Have details been worked out, such
as
whether it would be a TCP or UDP connection? Or was this intentionally postponed while T.130 work proceeded? If that's the case, should it be resurrected considering the state of T.130? Is H.281 simply considered
not
suitable for use on IP/Internet (due to latencies)? If so, is there
desire
to work out H.245 extensions to do the same things? I'm sure you've
all
discussed this before I was around; sorry.
-- Toby
Toby Nixon, Program Manager - NetMeeting http://www.microsoft.com/netmeeting Microsoft Corporation, Applications and Internet Client Group, Redmond
WA
USA +1 (425) 936-2792 Fax: +1 (425) 936-7329
mailto:tnixon@microsoft.com
Dave Lindbergh Email: lindbergh@pictel.com Manager, Technical Standards Group Voice: +1 978 623 4351 PictureTel Corporation Fax: +1 978 749 2804 100 Minuteman Road - M/S 635 H.320: +1 978 437 0265 (twice) Andover MA 01810 USA http://standards.pictel.com PGP fingerprint: C8 3E 28 69 53 0A 4B 87 EF E0 11 CD AC 87 4D CD
Dave Lindbergh Email: lindbergh@pictel.com Manager, Technical Standards Group Voice: +1 978 623 4351 PictureTel Corporation Fax: +1 978 749 2804 100 Minuteman Road - M/S 635 H.320: +1 978 437 0265 (twice) Andover MA 01810 USA http://standards.pictel.com PGP fingerprint: C8 3E 28 69 53 0A 4B 87 EF E0 11 CD AC 87 4D CD
------------------------------------------------------------------- Dave Lindbergh Email: lindbergh@pictel.com Manager, Technical Standards Group Voice: +1 978 623 4351 PictureTel Corporation Fax: +1 978 749 2804 100 Minuteman Road - M/S 635 H.320: +1 978 437 0265 (twice) Andover MA 01810 USA http://standards.pictel.com PGP fingerprint: C8 3E 28 69 53 0A 4B 87 EF E0 11 CD AC 87 4D CD -------------------------------------------------------------------