[Vorbis] Sample rates higher than 48k? 24-bit depth?

[Ian Malone]

Monty wrote:
> Higher sampling rates are not trying to provide more high-frequency
> headroom; they're designed to allow using less aggressive analog
> lowpasses during sample and playback.  The filters required by 44.1kHz
> and 48kHz sampling rates need to be extremely aggressive to avoid
> aliasing as they have so little room in which to roll off.  This leads
> to a very harsh filter that usually includes a plethora of ripple in
> the passband, rolling off too early, aliasing, or all three.  A 96kHz
> rate makes getting a good result out of the lowpass much easier (you
> have a full octave for rolloff instead of a tiny sliver of spectrum).
> 

So should a lossy codec worry about preserving high frequency
components?  Or would it be valid to discard them when encoding
and zero-pad the Fourier space data on decoding (basically a digital
filter)?

And for a bonus question: is it useful to up-sample (eg. to 96kHz)
before playback if the hardware can handle the higher rate?
(Assumes, I suppose, that the recording is done at a higher rate,
which I've been led to believe it often is for the above reasons).

-- 
imalone