Since the other responses seem to have gone off into a discussion of
firewalls, let me respond to this more directly by recalling a bit of past
discussion.
1) On the Megaco list, we finally agreed that the text vs. binary question
is something to be resolved by experiment.
2) In Turin (I think), the general opinion was that the protocol would have
to be binary to achieve the required performance. Interestingly enough,
however, no one that I recall felt that it should be PER ASN.1 (or even, I
think, BER ASN.1). Good reasons were given at the time. It would be good
if someone can remember them, but I think it has something to do with
processing performance and less risk of fragmentation than call signalling.
3) We have had an occasionally recurring suggestion on the Megaco list that
the Media Gateway control protocol should conform to the format of other
messaging out of the MGC, for the reason you give (i.e. to minimize
transcoding). Interestingly, the person advocating this was talking about
billing data, not signalling. The counter-argument on the list is that the
Media Gateway control protocol must work with different signalling protocols
(for example in H.GCP's case, with any of the H-series systems), so there is
no point in optimizing it for just one.
The real sticking point of debate is going to be whether the media
description structures will be those defined for H.245 OLC etc., or SDP.
Text vs. binary is a closely related but broader debate.
I'd suggest we start thinking about experimental design.
-----Original Message-----
From: Ami Amir [SMTP:amir@RADVISION.RAD.CO.IL]
Sent: Wednesday, March 31, 1999 4:15 AM
To: ITU-SG16@MAILBAG.INTEL.COM
Subject: H.320 gateways a MEGACO / ITU
Hi
During the last MEGACO conference call - Tom Taylor asked what could
hinder
ITU acceptance of the MEGACO work. It is obviously to everybody's interest
that IETF and ITU standards merge. As an example - it would be great if
the
MEGACO work could become an ITU SG 16 contribution.
An obvious item is efficient multimedia support (as reflected also in
John's
mail).
However, I think that one of the even more major difficulties facing us
will
be the encoding scheme (ASN.1 vs Text).
There are many who feel that ASN.1 is too heavy and complex for simple
devices, and should be avoided. This was one of the major reasons for the
emergence of SIP.
On the other hand, experience in the ITU H.323 work has shown that since
ASN.1 is the encoding scheme on the PSTN side, the use of ASN cleared the
way for PSTN to IP interoperability. This feature will be extremely
important in hybrid networks that need to provide Intelligent Network (IN)
services (e.g. "800"), while retaining the investment in existing IN,
billing and directory services (411).
Another problem is that if ASN.1 is not chosen, every device that will
have
to connect between a MEGACO component and H.323 will need to dis-assemble
and re-assemble (transcode) messages, and hence network performance will
suffer, and those devices will be more complex. A prime example - MGC to
GK
communications.
I am not promoting a specific approach. I just think that this complex
issue
needs to be addressed if we want to be able to be able accept a universal
protocol.
Do you think this is really a problem?
If so - any ideas on how to bridge the gap?
Ami
-----Original Message-----
From: John Segers [SMTP:jsegers@lucent.com]
Sent: Tuesday, March 30, 1999 6:27 PM
To: ITU-SG16@mailbag.cps.intel.com
Subject: H.320 gateways
People,
In yesterday's conference call, the subject of H.320 GWs was
raised
briefly. In my opinion, the connection model and protocol should
be
able
to deal with H.320. I would like to continue discussion on it on
the
mailing list.
H.320 allows a user to have a session with both audio and video on
a
single 64 kbit/s channel such as an ISDN B-channel. The same
channel
carries some signalling information (frame alignment, bitrate
allocation). To a MG supporting H.320, this means that on a
single
endpoint, three streams can come in, carrying different types of
media.
The current connection model of megaco/H.gcp does not cater to
this.
I
see two possible solutions:
The first is to allow multiple media in one context and to
describe
for
terminations the logical streams they carry. In a picture:
+----------+
| |
| +--------- signalling (FAS, BAS)
| |
B-channel ==========+ +--------- audio (16 kbit/s)
| |
| +--------- video (46.4 kbit/s)
| |
+----------+
The second solution is to have separate terminations for the
different
streams. They would all "connect to" the same physical endpoint.
In
order to properly identify the terminations, it is necessary to
have
logical names for them. The physical endpoint they connect may
have
the
hierarchical name proposed in the megaco document.
Another example of a H.320 session is the case of two B-channels
being
used for an audiovisual call. The following frame structure is
then
possible.
+--------------------------++-----------------------+
| Channel 1 || Channel 2 |
+-----+--+--+--+--+--+--+--++--+--+--+--+--+--+--+--+
|Bit 1|B2|B3|B4|B5|B6|B7|B8||B1|B2|B3|B4|B5|B6|B7|B8|
+-----+--+--+--+--+--+--+--++--+--+--+--+--+--+--+--+
1 | a1 |a2|a3|a4|a5|a6|v1|F ||v2|v3|v4|v5|v6|v7|v8|F |
2 | a7 |a8|a9|a |a |a |v9|F ||v |v |v |v |v |v |v |F |
3 | a |a |a |a |a |a |v |F ||v |v |v |v |v |v |v |F |
4 | a | | | | |a |v |F ||v | |v |F |
5 | a | | | | |a |v |F ||v | |v |F |
6 | a | | | | |a |v |F ||v | |v |F |
7 | a | | | | |a |v |F ||v | |v |F |
8 | a | | | | |a |v |F ||v | |v |F |
+---------------------------------------------------+
9 | a | | | | |a |v |B ||v | |v |B |
10 | a | | | | |a |v |B ||v | |v |B |
11 | a | | | | |a |v |B ||v | |v |B |
12 | a | | | | |a |v |B ||v | |v |B |
13 | a | | | | |a |v |B ||v | |v |B |
14 | a | | | | |a |v |B ||v | |v |B |
15 | a | | | | |a |v |B ||v | |v |B |
16 | a | | | | |a |v |B ||v | |v |B |
+---------------------------------------------------+
17 | a | | | | |a |v |v ||v | |v |v |
.
.
.
80 | a | | | | |a |v |v ||v | |v |v |
+---------------------------------------------------+
(a=audio, v=video, F=FAS, B=BAS).
We see that the video stream is split up over two channels. In
order to
cater to this, it seems we have to allow terminations to receive
media
from and send it to multiple physical endpoints. The two
approaches
outlined above can both be extended to allow this. Both extensions
will
lead to the introduction of logical names for terminations. In
the
first approach there will be one termination "containing" two
B-channels
on one side and three logical streams on the other. In the second
approach there will be three terminations, the one for the video
stream
referencing both B-channels, the ones for signalling and audio
referencing only channel 1.
The second approach allows us to keep separate contexts for
different
media types. It is then easy to delete, for instance, the video
part of
a session (session used loosely to desribe the contexts for the
audio
and video).
The first approach groups the streams coming from/going to one
user,
making it possible to remove a user from a context more easily.
Personally, I can't decide which approach I would prefer. How do
others
feel about these ideas?
Regards,
John Segers
--
John Segers email:
jsegers@lucent.com
Lucent Technologies Room HE
344
Dept. Forward Looking Work phone: +31 35 687
4724
P.O. Box 18, 1270 AA Huizen fax: +31 35 687
5954
Mike Buckley
Tom-PT Taylor