Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/3/84; site enmasse.UUCP
Path: utzoo!watmath!clyde!bonnie!akgua!whuxlm!harpo!decvax!genrad!panda!enmasse!mroddy
From: mroddy@enmasse.UUCP (Mark Roddy)
Newsgroups: net.micro.atari
Subject: Multitasking on the Amiga
Message-ID: <450@enmasse.UUCP>
Date: Mon, 19-Aug-85 09:39:33 EDT
Article-I.D.: enmasse.450
Posted: Mon Aug 19 09:39:33 1985
Date-Received: Sat, 24-Aug-85 14:14:11 EDT
References: <268@ihnet.UUCP> <1669@hao.UUCP> <5436@fortune.UUCP> <338@eneevax.UUCP> <2344@watcgl.UUCP>
Distribution: net
Organization: Enmasse Computer Corp., Acton, Mass.
Lines: 30

> Not true.  Programs can be arbitrarily large (since addresses are relocated
> at process startup time) and they don't have to be well behaved
> (since address faults, etc are trapped on a per-process basis).
> 

Well I for one don't see how either of these assertions can be true.

1) Programs can't be arbitrarily large, they have to fit into what is left
of memory, a variable dependent on what happens to be loaded at the time.
This is not the same as a multiprocessing system that allows any combination
of processes to coexist, limiting only the number of concurrent processes.

2) They do have to be well behaved. This is far more serious a problem with
the base register relocation scheme used in the Amiga. There is no way to 
prevent task_1 from writing all over task_2. Without the hardware protection
offered by an MMU, you are left with trust as the only protection between
the address space of individual tasks. I must assume that the address faults
mentioned above are writing off the end of known space type faults.

> I got this information from one of the people who designed the Amiga's
> multitasking operating system.  All in all, I was very impressed
> by the number of things that they got right.

These people are not objective.
-- 
						Mark Roddy
						Net working,
						Just reading the news.

					(harvard!talcott!panda!enmasse!mroddy)