Re: [rtcweb] RTCWEB needs an Internet Codec

[Ron <[e-mail]>]

On Fri, Aug 31, 2012 at 08:40:43AM -0700, Cameron Byrne wrote:
> Agree with the above. G.711 for mti only.
> 
> I want to push Opus and promote its use, but MTI is not the right method
> for that.

I haven't got the impression from this discussion that anybody thinks it
is about "pushing Opus" at all.  The question is what is best for rtcweb.

Since the charter of this group is to produce a standard that facilitates
"direct interactive rich communication", using things that already exist
as much as possible - I thought it was clear that the only important point
is that all this engineering work to produce rtcweb should not be a complete
waste of time for the people actually implementing it, and produce something
that gives such a poor experience in its baseline configuration that nobody
would ever actually want to use it for any of its intended purposes.

That Opus is a very obvious choice of existing technology, which no other
single codec comes anywhere close to matching for this purpose, stands out
like an elephant in the room.  Blind men might argue about whether it is a
pillar or a brush, or whether enough double blinded men have yet touched it
to be sure of this, but no amount of "pushing" is going to realistically
move it from that position one way or the other.

Pretending that other much-poorer codecs could fill this role is abandoning
the "rich communication" part of the charter, and condemning this group to
an embarrassing failure to achieve its goals in any meaningful way.

Let's not do that shall we.  It seems quite clear that many people have an
intense interest in its success.  Delaying this indefinitely, or selecting
poorer options than are already available today do not seem like the choices
that rational people who actually do want this to succeed would make.

Interoperability requires an adequate MTI codec.  Nobody has proposed an
alternative codec that is not clearly inadequate by comparison to Opus.
In a room full of clever engineers who want rtcweb to excel as a best of
breed specification, this decision should be a total no-brainer.

This time would be much better spent on the Actually Hard questions that
still remain for rtcweb.

 Best,
 Ron