[geeks] Kids today!

[Mike Meredith]

On Sunday 11 May 2003 16:57, Lionel Peterson wrote:
> --- Mike Meredith <mike at blackhairy.demon.co.uk> wrote:
> > On Sunday 11 May 2003 14:36, Lionel Peterson wrote:
> > > Mainframe O/Ss, like MVS (or it is now what, z/OS?) had a great
> > > feature, every time you issue a system call, every other process
> > > get the chance to execute.
> >
> > Read the end of page 21 of "The Design of the UNIX Operating System".
> > The Unix
>
> Well, a better reference would be "Blance DuBois - I've always relied
> on the kindness of strangers" - form a Tennessee Williams play
> (Streetcar named desire?)...

I'm a fan of Tennessee Williams but I suspect Maurice J Bach knows slightly 
more about Unix schedulers. Allowing other processes another chunk of CPU 
time every time a process does a system call is something UNIX could do (and 
certainly does when it comes to I/O), but more importantly UNIX will 
interrupt a process to switch to another after a certain amount of time.

Thus allowing a UNIX system to run numerous processes that are doing the 
equivalent of :-

	while (1) {
		;
	}

and still maintain a reasonable level of responsiveness. Actually that 
infinite loop should have been written in Ada which is where I've seen most 
examples of this happenning.

If the MVS scheduler works in the way that you suggest (and I strongly suspect 
it does not) then it will encounter a severe problem whenever it encounters 
an undergraduate learning Ada. Or any other naive programmer.

The UNIX scheduler has not relied on the "generosity of strangers" (i.e. 
co-operative multitasking) since the very earliest days (pre C).

> Yes, I know about "acct" - Ken & Dennis didn't sit down to think of a
> way to charge users for CPU time, they wanted to share - suits wanted
> to charge, and nothing says suits like a mianframe!

I'd have said that nothing says suits like trying to replace a mainframe with 
NT servers :)