LR> Does any similar definion exist within OS/2 for high 
LR> priority - aka "real-time" tasks?

I think Craig and Peter covered this well, but to add, think of OS/2 
"real time" tasks with a grain of salt.  In any true real-time
system,  you handle real-time events with an interrupt handler. It's
not kosher,  but I've written huge (200-400 clock) interrupt handlers
because I KNEW  what the other interrupt traffic would be on this
dedicated system. In  DOS, the typical way is to post a flag in the
handler, chain to any DOS  stuff hanging on the same int (which you
arrange to be nothing!), and  grab the return from the chain and
process your h/w stuff. An overrun  flag causes an error trap if you
get the same interrrupt again while  you're busy doing this. You'd want
to do LOTS of processing in C++ or better, optimized Pentium asm in an
OS/2 driver if you could; unlike  DOS, you might not be able to be so
cavalier about interrupt processing  time, even if you give up the
interrupt itself. I ain't never done it!

As for your comment on UNIX, my understanding is that ANY UNIX task
that happens to make a kernel call can tie up interrupt response and 
certainly processing by a task linked to the interrupt until the
kernel  processing completes, which is non-deterministic. there are
resulting a  plethora of unixes claiming "real time" capability and you
are then tied to them and/or the associated hardware for dubious
improvement. i think OS/2 at least lets the interrupt break thru.

___
 X KWQ/2 1.2e X See 7 states from Rock City!

--- Maximus/2 2.01wb