[xiph-rtp] Speex -02 feedback and m bit question.

[Phil Kerr]

Hi Greg/Jean-Marc,

You have probably seen Colin Perkins' feedback on the AVT list regarding
the -02 Speex draft. Having it move to WG status for the next revision
is great news!

The use of the M bit in the RTP header to indicate comfort noise is
unusual. Can you comment why it was used in this way, rather than
being used to indicate the first packet after the "silence" period,
as would be typical? The unusual semantics are not necessarily
problematic, but we I'd like to understand what benefit we gain from
the additional complexity of atypical usage.

From the draft we specify:

The M bit indicates if the packet contains comfort noise. This field
is used in conjunction with the cng SDP attribute and is detailed
further in section 5 below. In normal usage this bit is set if the
packet contains comfort noise.

..

cng: comfort noise generation - either 'on' or 'off'. If off
then silence frames will be silent; if 'on' then those frames will
be filled with comfort noise.

Should this really conform strictly to the RTP specs or is this a case
where the RTP transportation needs to bend towards the encoder?

I think his other points, fixing the p bit and moving the ITU related
text to the speex.org site are fair and should pose no problems.

Cheers

Phil

[Jean-Marc Valin]

Hi,

I haven't see the thread on the avt list, can you send me the link?

I'm not sure how the CNG is supposed to work. What I called cng in the
current draft is only about doing cng in the case of discontinuous
transmission (perhaps it could be renamed dtx_cng), so that when no
packets are sent, the decoder knows whether it should generate comfort
noise or just put zeros.

I don't know what the p bit is, so I can't comment on it. Same about the
"ITU related text".

Jean-Marc

Post by Phil Kerr
Hi Greg/Jean-Marc,
You have probably seen Colin Perkins' feedback on the AVT list regarding
the -02 Speex draft. Having it move to WG status for the next revision
is great news!

The use of the M bit in the RTP header to indicate comfort noise is
unusual. Can you comment why it was used in this way, rather than
being used to indicate the first packet after the "silence" period,
as would be typical? The unusual semantics are not necessarily
problematic, but we I'd like to understand what benefit we gain from
the additional complexity of atypical usage.

The M bit indicates if the packet contains comfort noise. This field
is used in conjunction with the cng SDP attribute and is detailed
further in section 5 below. In normal usage this bit is set if the
packet contains comfort noise.
..
cng: comfort noise generation - either 'on' or 'off'. If off
then silence frames will be silent; if 'on' then those frames will
be filled with comfort noise.
Should this really conform strictly to the RTP specs or is this a case
where the RTP transportation needs to bend towards the encoder?
I think his other points, fixing the p bit and moving the ITU related
text to the speex.org site are fair and should pose no problems.
Cheers
Phil

[Phil Kerr]

Post by Jean-Marc Valin
Hi,
I haven't see the thread on the avt list, can you send me the link?

http://www.ietf.org/mail-archive/web/avt/current/msg05323.html

Post by Jean-Marc Valin
I'm not sure how the CNG is supposed to work. What I called cng in the
current draft is only about doing cng in the case of discontinuous
transmission (perhaps it could be renamed dtx_cng), so that when no
packets are sent, the decoder knows whether it should generate comfort
noise or just put zeros.

Ok, this does sound like it can be handled by the standard RTP comfort
noise handling.

Post by Jean-Marc Valin
I don't know what the p bit is, so I can't comment on it. Same about the
"ITU related text".
Jean-Marc

Post by Phil Kerr
Hi Greg/Jean-Marc,
You have probably seen Colin Perkins' feedback on the AVT list regarding
the -02 Speex draft. Having it move to WG status for the next revision
is great news!

The use of the M bit in the RTP header to indicate comfort noise is
unusual. Can you comment why it was used in this way, rather than
being used to indicate the first packet after the "silence" period,
as would be typical? The unusual semantics are not necessarily
problematic, but we I'd like to understand what benefit we gain from
the additional complexity of atypical usage.

The M bit indicates if the packet contains comfort noise. This field
is used in conjunction with the cng SDP attribute and is detailed
further in section 5 below. In normal usage this bit is set if the
packet contains comfort noise.
..
cng: comfort noise generation - either 'on' or 'off'. If off
then silence frames will be silent; if 'on' then those frames will
be filled with comfort noise.
Should this really conform strictly to the RTP specs or is this a case
where the RTP transportation needs to bend towards the encoder?
I think his other points, fixing the p bit and moving the ITU related
text to the speex.org site are fair and should pose no problems.
Cheers
Phil