Re: a joint letter on low latency and Linux

[Dale Amon]

>There are two different problem spaces:
>1. Live -- which seems to me to be a hard realtime application when you
>    talk about serious quality.
>   Time_data_in - Time_data_out < K 
>  
>2. Places where some serious buffering (ok, not seconds, perhaps) is ok.
>   I thought we were discussing this -- and people here have been saying
>   that there is a great deal of applications in this space.

Since I've worked professionally in a couple of these areas
(live cybercasting and as a professional touring musician) I'd
define it slightly differently.

It isn't entirely a question of live vs nonlive. A video stream from
a live 'cast can just as well buffer. The fact that a viewing client
is 10 or 20 seconds behind real life isn't relevant.  

Likewise, as others have pointed out, taking a stream from a
CD, DVD or whatever can just as easily be buffered.

Where it all matters is in two cases that I can think of:

	* The mixing desk.
	* Performance.

A mixing desk has to be able to take multiple streams from
multiple sources, line them up and generate a glitch free
master. It has to be realtime because in the music biz this
is mostly done by ear (although there is also some queing
done as well, ie things like Cakewalk and other systems that
are already digital).

I would think that even a mixer can do a *small* bit of buffering,
so long as all tracks are kept in absolute sync and the worst delay
isn't enough to drive the tech behind the sliders over the wall. 
The closer to absolute tracking though, the more "pro" the kit.
In a world long ago and far away I occasionally ran the board
at some venue's when I wasn't playing myself. You've got such
long delays in the house anyway that it hardly matters if there
is a slight delay line. Sometimes this is done intentionally in 
the kit anyway: it's one anti-feedback technique. 

Studio's are far harsher of course.

Instrumental performance is the most demanding of them all. If
I am onstage in front of a live audience and playing an
improvisation with other musicians, I wonder if even
5ms is fast enough. Buffering is useless because if I
strike a note, the attack must be nearly instantaneous.
If you have never done live improv with really good musicians
you won't believe how tight this can be. [And if you've ever 
been on stage with bad sound trying to do this you know what
pure hell is.]

PS: I used to work in the Computer Music Lab at CMU in the 80's. 

------------------------------------------------------
Use Linux: A computer        Dale Amon, CEO/MD
is a terrible thing          Village Networking Ltd
to waste.                    Belfast, Northern Ireland
------------------------------------------------------