[Vorbis] Bitrate Peeling (no. really)

[ben-extra at MIT.EDU]

>Date: Fri, 1 Oct 2004 13:49:09 -0700
>From: Ralph Giles <giles at xiph.org>

>On Thu, Sep 30, 2004 at 03:41:18PM -0400, Monty wrote:
>
>> One approach seeks to implement a peeler that will work on any stream.
>> I think this approach is perfectly possible.  However...
>> 
>> Approach B relies on the encoder pre-structuring a stream to be easy
>> to peel.  I think this approach is simpler/faster to implement in the
>> decoder/splitter, and will result in higher final quality.  It will
>> require a more complex encoder.
>
>So, how do you think we should structure the bounty? One for each will 
>probably just divide the resources. One bounty for either?
>

>From my reading, there are two solutions that might be solved.  
A) works on existing streams
B) works on new streams only

I, for one, am *much* more interested in approach A - I adopted Vorbis
assuming this would soon be available.  But approach B would be
valuable.  Why not have two pots of money, since Approach A seems a
superset of approach B.

Pot one would be for approach A.  As an example, say I put $4 in it.
Pot two would be for approach B.  Suppose I put $a in it.

If someone solved A, they are by extension solving B, they get $5

If someone solved B, they would get $1.

If, later, someone developed a "re-organizer" for existing streams, such
that they could be peeled, they would get the remaining $4.

-Ben