Re: [rtcweb] RTCWEB needs an Internet Codec

[John Leslie <[e-mail]>]

Randell Jesup <[e-mail]> wrote:
> 
> If you're talking audio-only calls, I agree - but for audio+video, you 
> want a lot of bandwidth available for video.  If the user has 128Kbps, I 
> don't want 56K of that devoted to audio; I probably want 15-25K for 
> audio (which can give good wideband or better performance with Opus), 
> and the rest for video.  I probably wouldn't open audio up to 56 or 64K 
> (using Opus) below 300-500K of total bandwidth, because of diminishing 
> rate of improvement over ~25K.

   128Kbps is marginal for audio+video, yes.

   However, I don't believe it wise for us to try to standardize anything
about the audio-vs-video tradeoff.

   I will state my opinion that good audio plus one frame per second is
better in a conferencing environment than bad audio plus 128Kbps of
video; but I'd never impose that choice on anyone else.

   The point is: G.711 is suboptimal, yes, but not unusable.

   Our issue here is Mandatory-to-Implement. It is very important to
have at least one MTI audio codec. I could live with that being G.711,
because I trust the market to _actually_ implement others.

> Having a *good* low-bitrate (and adaptable) audio codec is *critical*
> to having a good solution for video.

   I'm not sure I agree, though it is certainly _very_ helpful.

> Without that all the cases that involve video are blown below 100's
> of Kbps; with a good low-bandwidth codec you can do video calls
> with 100Kbps or even less.

   I'm not sure even this will remain true for five years.

   But I am sure that the market will implement Opus or something
like it within five years without our needing to call it MTI.

> This is why I support Opus (+G.711 for legacy interop and testing)
> for MTI.

   This _is_ a good argument: I don't mean to ridicule it.

   But we need to reach closure on this. If enough people are scared
of the IPR questions of Opus, I'm willing to retreat to a SHOULD
implement Opus.

--
John Leslie <[e-mail]>