I don't know if I qualify as an "audio expert", but I have used the
audio facilities of three or four different workstations with sound
(Sparcs and 2 types of SGIs with totally different sound hardware; and
a Mac), written several tools for conversion between various sound
formats, and written several audio-using programs. My conclusions are
(1) sound with a header is better and (2) the sun/next format now
proposed as audio/basic is adequate.
Headers are important because with them you don't have to remember in
what format a file was recorded if it is separated from the mail
header. The more recent systems all have standard tools to record and
play sound files with headers, so shell scripts needn't worry.
While there are audio formats that are much more powerful than
audio/basic, their main usage is to store sampled voices for
instruments, which contain (for instance) a middle part that can be
repeated a variable number of times, or to record entire songs
composed of note references, etc. This is obviously overkill for
audio mail, where the basic requirement is to sample some audio and
play it back unchanged at a later time. The main differences between
available equipment are adequately represented by the fields of the
audio header: sampling rate, sample size, number of channels, encoding
type.
Cheers,
--Guido van Rossum, CWI, Amsterdam <guido(_at_)cwi(_dot_)nl>
"One's never alone with a rubber duck."