Path: utzoo!utgpu!water!watmath!clyde!att!ihlpa!ihnp4!cbmvax!daveh
From: daveh@cbmvax.UUCP (Dave Haynie)
Newsgroups: comp.sys.amiga
Subject: Re: Amiga UNIX
Message-ID: <4071@cbmvax.UUCP>
Date: 20 Jun 88 23:20:01 GMT
References: <23602@hi.unm.edu>
Organization: Commodore Technology, West Chester, PA
Lines: 38

in article <23602@hi.unm.edu>, erikj@hi.unm.edu (Erik Johannes) says:

> In order for the fork command to function properly the 68000 would have to
> chase down all the pointers and stuff, which would be extremely complex and
> time consuming.  

If at all possible, which it might not be.  Your fork routine would have
to identify somehow every data item that's a pointer.  No real way to do
this.  One alternate non-MMU method is to do a real CPU copy of the data to
the base location, which would achieve the same effect as the MMU, but be
far to slow for any practical use.

> One of the principal foundations of UNIX is cheap processes,

Compared to the above, perhaps.  But UNIX processes are certainly more
expensive than AmigaOS processes.  Which is why they created "threads"
in Mach.

> 	Intel got lucky with their convoluded segment registers in the 80x86
> family.  Since there is a Data base register, they can make a copy of the
> data segment and therefore have a true fork.  

If you wanted to restrict your 68000 system in the same way as the 80x86,
to 64K segments, you could allow only register relative code to be used.
This would give you a fork that's just as good as any on the 80x86 machines.

> For a 68k system, one must have an MMU or a 
> 68020 (68030) which has a built in MMU.

Only the 68030 has a built-in MMU.  Though the 68020 works quite nicely
with the  68851 PMMU.

> 
> 			-Erik Johannes
-- 
Dave Haynie  "The 32 Bit Guy"     Commodore-Amiga  "The Crew That Never Rests"
   {ihnp4|uunet|rutgers}!cbmvax!daveh      PLINK: D-DAVE H     BIX: hazy
		"I can't relax, 'cause I'm a Boinger!"