Re: [rtcweb] Making both VP8 and H264 MTI

[Ron <[e-mail]>]

On Thu, Nov 07, 2013 at 03:04:06PM +0900, David Singer wrote:
> 
> On Nov 7, 2013, at 5:33 , Mohammed Raad <[e-mail]> wrote:
> 
> > Hi,
> > 
> > There isn't much new in the arguments for and against VP8 vs H.264. Its
> > clear that there are concerns that are beyond how well each codec performs
> > and whether or not it is free. Although I recognize that the transcoding
> > option does not address the P2P use case, I do think that the goal of
> > interoperability can be better achieved through a two phase approach. The
> > first is enabling end-points to inter-operate through the availability of
> > transcoding from the service provider. The second is to make one of the
> > codecs the defacto MTI because of the its higher level of adoption. Yes, we
> > may never get to a single MTI if we follow this path but at least we will
> > have endpoints inter-operating. Besides, there are multiple other
> > difficulties in getting seamless P2P communications working all the time,
> > so I would suggest focusing on the service provider centered use case
> > first.
> 
> You know, you're right.  
> 
> We could suggest or even recommend both, and mandate you must implement at
> least one of the two. That would identify precisely two interoperability
> points in an otherwise much larger space, at least, and make it fairly simple
> for those that can and wish to, to offer transcoding, since the two ends are
> now defined.  It's not ideal, but it might be better than saying nothing at
> all.

I thought the performance and other problems with forcing _everyone_ through
a central relay server were well known?  I would have thought David at least
would be somewhat aware of them.  Adding that to a guaranteed interop failure
outcome might be a cunning plan, but I wouldn't call it a good one.

http://arstechnica.com/tech-policy/2013/08/report-after-patent-loss-apple-tweaks-facetime-and-logs-500000-complaints/

The whole point of a MTI codec is there will be no "otherwise larger space"
if we choose one that all implementors are free to use.

  Ron