Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!ll-xn!mit-eddie!zrm
From: zrm@eddie.MIT.EDU (Zigurd R. Mednieks)
Newsgroups: comp.sys.mac
Subject: Re: Mac Multitasking
Message-ID: <6611@eddie.MIT.EDU>
Date: Wed, 19-Aug-87 09:54:44 EDT
Article-I.D.: eddie.6611
Posted: Wed Aug 19 09:54:44 1987
Date-Received: Fri, 21-Aug-87 07:22:50 EDT
References: <6565@eddie.MIT.EDU> <2742@hoptoad.uucp> <3638@cit-vax.Caltech.Edu> <2758@hoptoad.uucp> <496@rocky.STANFORD.EDU>
Reply-To: zrm@eddie.MIT.EDU (Zigurd R. Mednieks)
Organization: MIT, EE/CS Computer Facilities, Cambridge, MA
Lines: 45

>
>I think I understand the Mac, so I will reply to this one...
>
>The multitasking we Amiga users talk about is the kind of 
>multitasking that sits there waiting for you to need it. It's not the
>kind of multitasking that the programmer or the user has to
>prepare for. Any computer (even the C64) can be made to print in the 
>background, but no programmer of user will take the time to do all the work,
>because it requires extra effort. Same with most Mac user's attitude towards
>multitasking --- They think it requires extra work, or means your programs
>will run twice as slow...
>
I think we are confusing multitasking with the relative merits of
premptive and non-premptive scheduling. While the Amiga may work well
with premptive scheduling, Unix, for example, does not work well on
low-powered micros. It is partly due to the fact that the Mac is a
single-tasking machine with limited (vertical-retrace manager)
background tasking that it can run a slick window system as well as it
does. It takes a 16Mhz 68020 to do as well with Unix and X or some
other windowing environment. Take a look at an old 68000 based Apollo
or Sun running some window system. It is pretty doggy.

So, in order to maintain the quality, not merely the capability, of
the Macintosh's interactive environment, non-preemptive scheduling
lets each task decide exactly how much of the processor it ought to
have. Badly behaved CPU hogs will, presumably, get bad reviews and
people won't buy them. That is probably a better way of weeding them
out than with some clever scheduling algorithm. Or, from another point
of view, I would rather see application designers making CPU usage
decisions than system designers.

Personally, I don't mind writing software that hooks into the
admittedly elaborate event processing structure of the Macintosh
because it gives me total control over what is going on. I won't mind
writing background tasks that use the various facilities in the
Macintosh for background tasking because they too will allow more
complete control over CPU allocation than premptive scheduling.

-Zigurd
----------------------------------------------------------------------
Zigurd Mednieks
MURSU Corporation
P.O. Box 1894
Cambridge, MA 02138
(617) 522-9035