-----Original Message-----
From: Colin Perkins [mailto:csp@csperkins.org]
Sent: Monday, April 26, 2004 10:48 AM
To: Gunnar Hellstrom
Cc: IETF AVT WG; Magnus Westerlund
Subject: Re: [AVT] Reminder: WG Last Call:
draft-ietf-avt-text-red-02.txt
On 11 Apr 2004, at 06:46, Gunnar Hellstrom wrote:
There has been a good discussion in the avt list about the use of the
a=fmtp parameter for text-red. The discussion has been linked to the
parallell work with real time text transmission in RFC2793bis, but is
valid
for text-red and should possibly influence the text-red draft
slightly.
Here are my review of text-red-02 including the result of the
discussion on
a=fmtp:
1. In section 4, there is this sentence:
"Any dynamic payload type in the list MUST NOT include its content-
type only the payload type number."
The "type only" comes out a bit strange. I suggest rewording to what
you
shall do:
"Any dynamic payload type in the list MUST be represented by its
payload
type number and not by its content-type."
okay
2. The offer-answer aspects of the a=fmtp parameter
In a separate mail in the avt thread discussing a=fmtp for RFC2198, I
enter
offer-answer considerations for the a=fmtp parameter and suggest that
this
sentence is included in the sdp mapping section of text-red spec:
"In an offer/answer situation, the fmtp parameter in the answer MAY
describe a different number of redundant levels than in the offer.
The
session SHOULD use the highest number of redundant levels described."
I don't think this is appropriate due to the problem of asymmetric
loss
we've previously discussed. I would suggest that the "text/red"
registration remain silent about this issue, and that we consider
offer/answer considerations for redundant audio when RFC 2198 is
revised for Draft Standard. This will ensure that we have consistent
specifications for all uses of the format.
--
Colin Perkins
http://csperkins.org/