Nathaniel Borenstein:
Bear in mind that in the computer world, u-law is to audio what ascii is
to text: flawed, not completely adequate, easy to complain about, and
nearly universally available. That we should have something like
audio/basic for simple interoperability strikes me as hard to argue
with. That there should also be a group of people designing a better
high-level audio subtype with higher fidelity & better compression is
also something I strongly believe. I'm not enough of an audio expert to
be a part of such an effort myself, but I'd love to see it happen, and
I'd hate to see that worthy goal set back by efforts to tweak
audio/basic in relatively minor ways (e.g. to get marginally more
compression, at the cost of a far less standard format....)
I'm sorry, but I would like to take exception with almost all of that!
First, U-LAW isn't nearly as universal as you believe. Most PC
platforms used to have 8-bit linear audio and now they're all
upgrading to 16-bit CD-quality. The SGI Indigo has CD-quality audio
hardware but only limited software support for playing U-LAW files.
Perhaps your perception of "universal" is somewhat distorted by having
used sparcstations too long?
After having listened to large quantities of U-LAW encoded music that
we transmit locally on the ethernet (an entertainment service called
"radio") I can hardly see what's so flawed about it -- it gives you
good AM-quality sound.
Halving the size of audio data isn't my idea of "marginally more
compression". The ADPCM format I'm proposing is in fact compatible
with Intel's audio format for DVI, one of the more popular multimedia
platforms available (and the availability of a *free*, fast
implementation means that Intel can't get a commercial advantage out
of that).
--Guido van Rossum, CWI, Amsterdam
<Guido(_dot_)van(_dot_)Rossum(_at_)cwi(_dot_)nl>