Path: utzoo!attcan!uunet!tut.cis.ohio-state.edu!zaphod.mps.ohio-state.edu!mips!apple!agate!stew.ssl.berkeley.edu!ericco
From: ericco@stew.ssl.berkeley.edu (Eric C. Olson)
Newsgroups: comp.sys.atari.st
Subject: Re: Unix on the Atari ST
Summary: Is it a good idea?
Keywords: multitasking multiprocessing os
Message-ID: <1990Jul1.174840.7379@agate.berkeley.edu>
Date: 1 Jul 90 17:48:40 GMT
References: <90176.164228RXS12@psuvm.psu.edu> <a7647881@mist.UUCP>
Sender: usenet@agate.berkeley.edu (USENET Administrator;;;;ZU44)
Organization: University of California, Berkeley
Lines: 48

I like Un*x.  I like my Sparc 1 at work.  I like X.  I like NFS.  Too
bad that Laser C on the Atari provides a comparable compiling
environment AND is faster for development.

I don't know that exact nature of the speed problem on the Sparcs, but
I think that its the multi-processing environment in Un*x.  Start up
times for processes are measure in seconds, since the kernal must
create a new virtual machine for each process.  The performance of
such a system makes it unreasonable to associate a single window with
a single process -- the time cost is too high.

I recently used a flight simulator on a Sparc -- incredibly
rudimentary, just like stuff you see on pcs (BTW, it also used a
graphics accelerator board).  I think its slow because to the
multi-processing philosophy -- the sever gets the mouse events,
figures out who owns them, sends them off, some computations are done,
a bunch of graphic primitives are sent back, which the sever displays.
Meanwhile, various processes are being swapped in and out memory.  The
bottom line is that the sparc machine is much more powerful than pcs,
but its user interface performance is not.

I've also used Lisp Machines and Xerox Interlisp machines.  These guys
are multi-tasking, and the interface has a much higher throughput.
Multi-tasking is coming back in vogue, with top people at Sun
indicating that they are interested in moving in this direction.
There is much discussion about light weight processes.

For the Atari ST, multi-tasking is really the only choice since it
can't provide the virtual machine environment required for
multi-processing.  I suspect the OS/9, Minix, and MTcsh are all true
multi-tasking OS.  I expect that no one would be able to tell the
difference between multi-processing and multi-tasking code that was
well written.

Of course, there is the system resource problem in multi-tasking --
usually memory violations in poorly written code.  But, the underlying
implementation of MTcsh seems like a very reasonable start at sharing
system resources such as disk drives, etc.  Use of a sophisticate
language like Lisp would eliminate memory violations.

In short, I think that there is a place for machines like the Atari ST
in the future.  But developers (and users) need to focus on computing
in the 90s, not what was in fashion in the 80s.   Light weight
processing in a robust environment is certainly an obtainable goal for
the Atari ST.

Eric
ericco@ssl.berkeley.edu