Path: utzoo!yunexus!unicus!craig
From: craig@unicus.UUCP (Craig D. Hubley)
Newsgroups: comp.sys.amiga
Subject: Re: Amiga IPC should be the Amiga Unix IPC!
Message-ID: <2536@unicus.UUCP>
Date: 15 Apr 88 00:46:50 GMT
Article-I.D.: unicus.2536
Posted: Thu Apr 14 20:46:50 1988
References: <2504@unicus.UUCP> <655@imagine.PAWL.RPI.EDU>
Reply-To: craig@Unicus.COM (Craig D. Hubley)
Organization: Unicus Software Inc., Toronto, Ont.
Lines: 59

In article <655@imagine.PAWL.RPI.EDU> jesup@pawl2.pawl.rpi.edu (Randell E. Jesup) writes:
>In article <2504@unicus.UUCP> craig@unicus.UUCP (Craig D. Hubley) writes:
>>  So that AmigaDOS tasks can talk
>>to Amiga Unix tasks (remember AmigaDOS will run as a task under Amiga Unix).
>
>	If I were asked how to implement Unix and still have AmigaDos, I
>would run Unix as a AmigaDos task.  Unix programs don't go to hardware
>directly, they go to the kernal.  Since C= is porting the kernal, it could go
>to the libraries (graphics, etc) for it's manipulations.  It is much easier
>to run a non-realtime system under a realtime one than vice versa.

Yow, come to think of it, you're right.  It *would* have to be the other
way around.  But then, that wouldn't be quite so difficult.  I don't
remember who said it originally, but perhaps I got it backwards.
However, one way or the other, having them both running at the same time
is *not* a `nice feature', its a necessity.  I don't want to go through
the kernel and up again to access the blitter, etc... it'll be as slow
as a Sun.  All I want is an AmigaDOG user interface using Unix as its
processing back-end.  That is, I don't want an improved Intuition under
Unix, because it will be slowwwww and many of the Amiga's real neat
features will go to waste... unless I misunderstood the Amiga Unix concept.
Anyone who knows what the interface will be like, please comment.

>then it could kill unix real fast when a program goes south.

This is a risk that presumably those that need lightweight multitasking
are willing to take in pursuit of their fanatical aims.  And a machine
going south once in a long long while (after you've got the bugs out)
is, in some environments, an acceptable hit to take for lightweight
multitasking.  I don't care if AmigaDOS goes, I just don't want the
obnoxious `multitasking is just like multiuser' UNIX mentality that
makes several closely cooperating tasks treat each other like lepers.
If someone has found a way to make Unix tasks lightweight, fine.
I expect we will have to wait until V.4 and then buy a SPARCintosh to see it.
Better start saving now.

In the meantime, I'll go looking for a decent *workstation* operating
system, and talk to Unix only at a distance.  Too bad, AmigaDOG could
have been it.  Unfortunately nobody took the time.  As an example of
what can be done, Eric Haberfellner's Handshake program is actually four
cooperating tasks (one waiting for serial input, one waiting on keyboard input,
one is the state machine, one blinks the cursor and other blinking text).
At least, that's what I gathered from his explanation at the Toronto ADF.
He said that the overhead for these *four* tasks was negligible, under 5%
of utilized CPU time for the whole program.  Imagine that in Unix.

It's stupid go back to faking multitasking in my own programs because 
the OS's multitasking overhead sucks shit, Unix or not.  I hope that
AmigaDOS doesn't die, because then I'd have to look at the only remaining
multitasking microcomputer operating system: *bluech* OS/2.  A sad fate.

Single-user multitasking != Multi-user multitasking

>     //	Randell Jesup			      Lunge Software Development

	Craig Hubley, Unicus Corporation, Toronto, Ont.
	craig@Unicus.COM				(Internet)
	{uunet!mnetor, utzoo!utcsri}!unicus!craig	(dumb uucp)
	mnetor!unicus!craig@uunet.uu.net		(dumb arpa)