Errors / ambiguities / problems found in specs at last week's int erop

[Paul Long]

Reinhard,

I have responded to each of the items you listed.

- I've wondered myself why there is a separate codepoint for G.729a. From my
understanding, the use of G.729 or G.729a is a local issue and does not
change the bits on the wire. Maybe we need to state in a future version of
H.323 that they are equivalent with respect to signaling.

- Are you saying that a GK cannot return a TTL in RCF that is larger than
the TTL in the corresponding RRQ or that it cannot subsequently increase the
TTL without first having received a lightweight RRQ? I believe that a GK
could return a larger value, but you are correct that there is no way for it
to update this value without responding to a lightweight RRQ.

I don't see how the mandatory-parameters issue is related to TTL.
Regardless, this points out a general weakness in semantics for later
additions to H.245 (since H.245v2) and especially H.225.0. IMO, the
description of all SEQUENCE OF and SET OF components should include the
semantics for these situations, even if it is "obvious" to the author what
the semantics should be:
    - OPTIONAL component whose encoding is absent
    - zero components (empty)
    - exactly one component
    - more than one component
    - component order (we should also use SEQUENCE OF and SET OF where
appropriate)

We continue to have interoperability problems due to the differing
interpretations as to what these mean.

- Regarding call reference value, since H.323 is the top-level
Recommendation that references H.225.0, if there is a conflict, H.323 is
always normative. This is true in all referencing relationships between
Recommendations, e.g., between H.225.0 and Q.931.

- I can't find in H.323 where destinationAddress is required. H.225.0v2 does
require it, though, "if available" (whatever that means).

- The audio framing expressed in H.245 is _not_ ambiguous. It is clearly
stated that this is the maximum framing. B.2.2.6/H.245v5: "When an H.225.0
multiplex is used, these numbers indicate the maximum number of audio frames
per packet." The problem is that some _implementations_ do not support
variable framing, e.g., some DSPs used in GWs can only process 30ms frames.
Possible solutions: "fix" decoders so that they support variable framing,
reframe into fixed-size frames before passing along to decoder, add ability
to express fixed framing in H.245.

- MCU prefix. I surprisingly have nothing to say about this one. :-)

- What is a "BE message format?"

- CRV. Wasn't this already described in a previous item?

Paul Long
Smith Micro Software, Inc.

-----Original Message-----
From: Reinhard Scholl [mailto:Reinhard.Scholl at etsi.fr]
Sent: Wednesday, March 01, 2000 10:34 AM
To: 'h323implementors at imtc.org'; 'ITU-SG16 at MAILBAG.INTEL.COM';
TIPHON_WG6
Subject: Errors / ambiguities / problems found in specs at last week's
int erop



The following errors / ambiguities / problems were reported at last
week's
ETSI-TIPHON/IMTC interop event at ETSI:


- Terminal capabilities for G.729 and G.729 Annex A. Though mutually
compatible, TCS often fails if vendors only specify support for one or
the other. E.g. 729 <--> 729a should work but TCS fails.

- H323: GK cannot increase TTL in RCF. Mandatory parameters of type
SEQUENCE OF ... can have empty list.

- H.225 says that call reference value should be same in ARQ and SETUP.
H.323 says that call reference value should be different in ARQ and
SETUP.

- In H.323, the destination address in UUIE in SETUP is mandatory, in
H.225.0 it is not.

- H.245: if a capability being sent includes G.711 "20 msec", the other
side may think that the sender supports both "20 msec" and "10 msec"
rates. But it can be that the originator supports only 20 msec.
Ambiguity.

- MCU should probably also have "prefix", not just the gateway.

- Need BE message formats so that packet dumps can be quickly
interpreted.

- When using a GK, should the CRV in the admission and setup be the
same?


        Reinhard Scholl
        ETSI