> On Mon, 9 Jun 1997 01:48:55 -0700 (PDT), Trent Piepho wrote:
>
> >On Mon, 9 Jun 1997, Colin L. Hildinger wrote:
> >> I like OS/2's method, personally. There are four brackets of priority,
> >> each with higher importance. A critical priority always takes
> >> precedence over an idle one. Period. If a critical priority thread is
> >> due, a server level one and below waits, if an server level thread is
> >> due, an application level and below waits, and if an application level
> >> thread is due, idle waits (I might not have gotten the names right).
> >
> >There are some flaws with a strict priority system like this. Say an
> >application program has locked a file that a server process needs. The
> >application will never get any cpu time and will never unlock the file. Thus
> >it is effectivly preventing the higher priority server process from running.
>
> I may of course be slightly wrong then. I know Idle priority works
> this way, the others may not be quite as strict, I know that idle is,
> though.
>
>
> Colin L. Hildinger
I spent a semester in an Operating Systems course, and we spent
months over different type of deadlock
prevention/detection/termination schemes. Finally, at the end, we
asked the instructor what kind Unix uses. He said "None. They all
use way too much CPU." So, if an app grabs a system file(which, by
definition it shouldn't ever do)and locks up the system, give if the
three finger salute.
-- Richard Hendricks, Applications Engineer, Austin, Texas