[ On Sunday, December 8, 2002 at 16:21:47 (-0500), Jed Davis wrote: ]
> Subject: Re: Scheduler hints (Re: CVS commit: syssrc/sys/kern)
>
> On Sat, Dec 07, 2002 at 12:59:13PM +1100, Daniel Carosone wrote:
> [snip]
> >  - moving the offender to the end of the run queue (from christos)
> 
> This seems like a pretty good idea to me

No, it's not really.  There's no way to prove the process which bumps
into the limit is an "offender", especially not from inside the kernel
where there's no way to define external global usage policies.

Indeed there's not necessarily anything wrong in such a situation.

Provided that the application does the correct thing to deal with its
inability to fork() at the moment then all's well.

The kernel's job is simply to schedule the processes and to run them all
as fast as it can and when a slot frees up to allow another fork() call
to fill it again.

-- 
								Greg A. Woods

+1 416 218-0098;            <g.a.woods@ieee.org>;           <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>