Re: [dispatch] [AVT] Proposal to form Internet Wideband Audio Codec WG

[Henry Sinnreich <[e-mail]>]

>at least people from the potential contributors (at least Skype, Speex, SPIRIT) are already pretty-good

As a user of all three codecs and some other as well:
The seem to me clearly among the best technology and in wide-spread use as well.

This should put to rest the questions of expertise, quality and availability of contributors and reviewers.

Henry


On 6/2/09 12:33 PM, "Slava Borilin" <[e-mail]> wrote:

I do not beleive the one that will come wil be low quality.
at least people from the potential contributors (at least Skype, Speex, SPIRIT) are already pretty-good in the commercially exploiting their own codecs on the market.
i think this is probably false alert.

regards,
Slava Borilin



________________________________
From: stephen botzko [mailto:[e-mail]]
Sent: Tuesday, June 02, 2009 9:32 PM
To: Henning Schulzrinne
Cc: Roni Even; [e-mail]; Jason Fischl; [e-mail]; [e-mail]; Slava Borilin
Subject: Re: [AVT] [dispatch] Proposal to form Internet Wideband Audio Codec WG

>>> - the quality of the codec may not be competitive

I think its very important that the codec quality be competitive.  People expect excellence from IETF standards  Standardizing non-competitive codecs because they are cheap does not seem to be a good choce.

Steve B.

On Tue, Jun 2, 2009 at 12:15 PM, Henning Schulzrinne <[e-mail]> wrote:
I  view this as a trade-off. If we pursue this, there are risks:

- nothing  may come of it since there are no experts willing to help
- somebody will  claim IPR on the resulting work
- the quality of the codec may not be  competitive

However, if we don't do this, we are stuck with the status  quo, which is not all that satisfactory. Thus, unless there are significant  costs for "innocent bystanders", I see this as a risk worth taking. In the  worst case, we are no worse off than we are today. In all other cases, we'll  have an additional choice for a wideband codec, even if it's not "the best",  just "good enough". After all, most people use G.711 today, which has a really  hard time making that claim.

Most real work in the IETF is done by very  small teams, typically less than 10, so as long as we have a handful of people  that are willing to contribute, this can work. It might even work better,  since you may get fewer people who have half-baked opinions - we may skip the  binary vs. XML debates...

We can set some ground rules ("must be tested  with packet loss of 5%") and then see what happens. Compared to most network  protocols, the likely negative impacts (such as security or congestion control  issues) of even a badly-designed codec are pretty limited.

Henning




On Jun 2, 2009, at 11:57 AM, Roni Even wrote:


Hi,
I  do not want to sound like someone who is for IPR (I am not), but why  stop
at codec, let's require it for all IETF work. There are IPR on IETF  work
which is must simpler, in my view, than wide band audio  codecs.

I think that we can start with "royalty free" even though I  am not sure that
it will accepted as part of the charter of any other  work group so why pick
on codecs which require more work.

This  leaves the other reasons I heard for doing it at the IETF which is  the
price of participating (cheaper than being an ITU-T member) and maybe  design
less expensive characterization  tests.

Roni


-----Original Message-----
From: Jean-Marc  Valin [mailto:[e-mail]]
Sent: Tuesday, June 02,  2009 6:14 PM
To: Roni Even
Cc: 'Slava Borilin'; [e-mail]; 'Jason Fischl';  [e-mail];
[e-mail]
Subject: Re: [AVT] [dispatch]  Proposal to form Internet Wideband Audio Codec
WG

Hi,

Roni  Even wrote:


I  am not sure what prevented you from doing it today at the ITU or
MPEG,  why do you see the IETF handling it differently.

I would also like  to remind you and Jean-Marc that once you are
bringing work to a  standard body it may require collaboration with
other people who will  have other proposals that will also address the
same requirements and  you may need to invest money in comparative
testing by independent  listening labs.

I also think that you will need to supply the codec  in source code and
provide copy right to the  IETF.


I am well aware that bringing work to the IETF  would require
collaboration with others. I am not seeking control over  the work I am
currently doing and would really welcome such  collaboration. The idea is
only to have the best wideband codec possible  without IPR issues. Given
the ITU and MPEG track record, I think it would  be very unlikely for any
of those organisations to work on an IPR-free  codec.

I also agree with Henry that "the Internet has different  criteria than
ITU-T networks may have". Internet adoption follows  different patterns
than adoption in the ITU primary target markets. For  example, the
Internet has more consumer-reconfigurable software, while  the ITU has to
deal with a lot of fixed hardware deployments. At the ITU,  it makes
sense to invest large sums of money into testing and  characterisation of
codecs because the codecs deployed there usually stay  around for a long
time and the infrastructure investments are usually  very large. On the
other hand, I would say the investments in codecs for  the Internet are
comparatively smaller and, while testing is still  important, it is not
as critical as it is for the ITU.

It's also a  choice one has to make. It is unlikely that companies would
invest money  in expensive testing if they are not going to obtain
royalties in return.  However, I think we may be able to define some more
lightweight  (collaborative?) testing that is sufficient and doesn't
involve as much  investments as what the ITU does. For the Internet, I
believe an IPR-free  codec that everyone agrees performs well is better
than a  patent-encumbered codec that has had more extensive testing. This
is  again another difference with the ITU: patent-encumbered codecs tend
to  hurt a lot more on the Internet because many applications (e.g.
giving  away the client) are very hard (or impossible) when one has to
pay  per-channel royalties.

As for the source code issue you pointed out,  all the Xiph codecs are
already published under a very permissive open  source license (BSD), so
this would not really change.

   Jean-Marc


Roni



*From:*  [e-mail] [mailto:[e-mail]]  *On Behalf
Of *Slava Borilin
*Sent:* Monday, June 01, 2009 11:50  PM
*To:* [e-mail]
*Cc:* [e-mail]; Jason  Fischl
*Subject:* Re: [AVT] [dispatch] Proposal to form Internet  Wideband
Audio Codec WG



Agree with Jean-Marc.
SPIRIT  is interested to contribute as well - having a dozen of  proprietary
codecs developed, including

one  specifically desgined for internet (SPIRIT IP-MR, which is now  under
WGLC on draft-ietf-avt-rtp-ipmr-04) -

multi-rate,  scalable, adaptive, wideband codec.

We can also continue this work  with IETF.

Moreover, most if not all efforts coming from ITU on  codecs unfortunately
are NOT really focused on

internet-specific  codecs (that's why several companies have had to invent)
-  as ITU preference is mainly

specific  (i.e. cellular) networks at first.

however, as we see the greant  rise of pure "internet-basd communications"
(skype,  webex/citrix, and many others) -

we  all (and users) are suffering from inefficiency in all  currently
"standard" codecs and ambiguity in the choice  of

internet-targeted  ones.

Again, we probably can put together enough number of  contributors to the
WG to have the expertise.


regards,
Slava  Borilin

--
John Lazzaro wrote:

  A traditional  response to this type of request is to note that the
IETF


   really doesn't have much in the way of expertise in audio  codec
design.


   I don't see many of the regulars who present at the AES codec  paper
sessions


   posting here on AVT (ditto ICASSP paper sessions for voice  codecs).
It's


   a full-time job to keep up-to-date and contribute to that

   signal-processing lore.



Well, there's no reason that the  IETF cannot build such an expertise
in audio codecs. This is actually  something in which I'd be interested
in getting involved and I'm sure  others at Xiph.Org would be
interested as well. We have several people  with audio codec expertise
from working on Vorbis, Speex and (more  recently) CELT. It turns out
that the CELT codec currently under  development at Xiph actually meets
most of the requirements from the  original proposal in being a very
low delay codec with adaptive  bit-rate and sampling rate (up to 48
kHz), scalable complexity, and  good robustness to packet loss. We'd be
willing to continue the  development with the IETF. Even if not with
CELT, it's still like to be  involved in such a new  WG.

Cheers,

 Jean-Marc

_______________________________________________
Audio/Video  Transport Working Group
[e-mail]
https://www.ietf.org/mailman/listinfo/avt


_______________________________________________
Audio/Video  Transport Working Group
[e-mail]
https://www.ietf.org/mailman/listinfo/avt