Date: 19 Jul 1994 16:55:19 U
From: "jbabcock" <jbabcock AT netrix DOT com>
Subject: RE: RE: Real-time OS under DJGPP
To: "djgpp" <djgpp AT sun DOT soe DOT clarkson DOT edu>,
        "Stephen Benson" <stephenb AT scribendum DOT win-uk DOT net>

>A better way to classify multitasking systems is to distinguish them as
>cooperative and preemptive....
....
>A more detailed description of the requirements for preemptive multitasking
can
>be found in most college level operating systems textbooks.

I think the point being made was that if a process relinquishes control of
processor and resources for any period of time, by definition it is not
functioning in "real time". 

-- 			              
+- Stephen Benson -+- -+- -+- -+- -+- -+- Linux 1.0.9/1.1.18  XFree86 2.1.1 -+
+- -+- stephenb AT scribendum DOT win-uk DOT net -+- -+- CI$ 100273 DOT 123 AT compuserve DOT com -+
                                       *.                                    
+- -+- -+- -+- -+- -+- "a microscopic cog in his catastrophic plan...." -+- -+
_______________________________________________________________________________
All real time systems have a certain latency associated with their response. 
Even an interrupt driven system has a measurable latency due to the saving of
the processor context prior to the ISR being invoked.  As has been pointed out
in other messages, the accepable latency is the driving factor for a
multitasking design.  By using a preemptive design instead of a cooperative
design, the latency is not dependent on the efficiency of the other tasks
running in the system.  When a priority system is included with the preemptive
design, the maximum latency for a high priority task can be guaranteed even
though several other tasks may be running in the system.  Amazingly enough,
this will provide a better response while getting more done than a single
threaded application.

I hope that I haven't muddied the waters too much here.

-Jim.