🚀 go-pugleaf

In article <41hubn$o3c@sundog.tiac.net>,
Jim Williams  <williams@tiac.net> wrote:
 >tzs@u.washington.edu (Tim Smith) wrote:
 >>
 >>Basically, optimal scheduling requires taking into account the *global*
 >>state of the system.  Cooperative systems make each process, which only
 >>knows the *local* state of the system, make scheduling decisions. 
 >>
 >
 >Well put!  As I have said before, cooperative systems make sense in
 >certain circumstances.  In the terms you've used above these cercumstances
 >are when the local state of the system for each process in the system
 >is known to each other process in the system.
 >
 >Cooperative systems do not work well in heterogenious situations.  The
 >programmer has to have complete control of the machine in order to
 >make cooperative systems work.

Also well put. Cooperative light-weight threading, for instance, _can_ be
made to work effectively. That's because the processes that comprise
the application are designed to work tightly together. From time to time
I work with a threading package that _used_ to be cooperative. The author
did have the good sense to convert it to pre-emptive mode. He had the good
sense to realize that it isn't that great even for LWP. Using cooperative
threading for an entire OS is an act deserving of corporal punishment.
-- 
Finger c2a192@keats.ugrad.cs.ubc.ca for PGP public key. Or contact
one of the public key servers -- the key id is 0xD3C7995D.