Sorry for the very long mail, I had been waiting too long... actually, I
forgot something :) My third point was about better audio processing:
I've recently been playing with replay-gain support in liquidsoap. I
won't detail that too more -- but it was a nice success. That made me
realize that it was much better to use some power measurement reflecting
how the human ear actually perceives each frequency (the replaygain
standard relies on that), instead of giving each frequency the same
weight by computing a simple RMS (that's liquidsoap currently).
I think it would be a great field for improvement in general: get more
serious audio processing. The best would be to get new people involved
with some background in audio. But we already have good devs, like
Balbinus, whose work on filters is still waiting for tiny adjustments in
order to be usable at best.
In the case of power measurement, it also calls for some nice design
allowing skip_blank() or smart_cross() operators to be parametrized by
different possible volume computations: RMS, human ear, dog ear, etc.
There was a hope to achieve that with interactive_floats but I think
it's dead: we should find a good solution here.
Enough feature requests!
Good night,
David
