Let's go for it then...

Hi there!,

    I'm very glad to hear from Yves Dhondt, Oztan Harmanci, Lukasz Tomczykiewicz et. al
that they are interested in joining the project.  May be this time we can give the definitive
push to the project.

    I think that we all agree that the reference software should be used as a test bench,
learning tool and cut&paste source, but that a new clean implementation shall be started.
I now realize that implementing it in C++ could be a problem.  Most of the people do not
know the language and many of them are do not feel comfortable with OO.   Then the
codec should be written in C, with the critic parts in MMX/AltiVec/3DNow! ASM code.
(OK, i initially proposed Java, then i resigned to use C++ and now C... don't even think
of it, i will not accept a Cobol implementation ;-)

    But even if it's developed in C i strongly believe that a fundamental feature must be
to support "plugable" modules.  That is, detailed interfaces must be defined for each
core process of the standard (transform, quantize, intra...) and should be possible to
interchange implementations very flexibly.  Of course this will come by the use of
function pointer tables (Xvid does something similar, no?).  I agree with the idea of 
Oztan of implementing suboptimal encoding processes for streaming like applications,
and mostly optimal code for offline processing (DVD ripping?).  I think that the plugable
modules are the key to achieve this seamlessly.

    So, we need to think about the basic structure to the program and organize a little
bit...

    /AITOR