🚀 go-pugleaf

>I'm sorry Matt, but I beg to differ.  I'm just plain sick and tired of
>having to reboot my machine just because of a wayward pointer in a program
>I'm developing.  SOME sort of memory management to protect myself from myself
>would be terrific.  An MMU doesn't have to be incompatible with the OS.  You
>could always disable it, but I think that for the extra protection it affords,
>it's a must.
>
>A multi-user system without an MMU is suicide!  I wouldn't buy it.  Not for
>my business, anyways.
>
>And you don't need an '020.  An '010 would do nicely.  Also, according to the
>Byte article on 68K MMUs, you can even get some MMUs for the 68000.

	Buy yourself a SUN.  I am being practical.  If C-A put a MMU in
their Amiga(s), the cost of the machines would be too high (in my opinion).
Besides, if your going to have an MMU, you also want full resource tracking
in the kernal so when programs die, they don't effect the rest of the 
machine.  Additionaly, the current message passing system assumes shared
memory, and that would have to be changed.

	In short, to add an MMU would require a complete rewriting of the OS
almost from scratch... If C-A stays with the current kernal model (which is
very well done by my standards), an MMU would be impractical.

	Then again, it is true that you could selectively enable an MMU for
specific programs which stay within certain bounds, if it weren't too
expensive, the best route to go is make it a programmer's option and allow
him to turn it on and off at will.


					-Matt