What the heck, lets go out on a philosophical limb :-)

Excerpts from mail: 26-Feb-98 Re: PowerMac types.h lossage
sakamoto@cec.co.jp (932*)

> <Pine.A32.3.96.980225110200.129195A-100000@macro.stanford.edu>
> skippy@macro.stanford.edu wrote:

> > > (someone else)
>  > > I wanted to solve this by creating sys/cpu/{m68k,powerpc}, but there were
>  > > some interesting issues revolving around exporting header files, etc.
>  > 
>  > What were the problems? Couldn't userland just have both cpu/powerpc and
>  > arch/powerpc go into /usr/includes/machine, or was that not the problem?
>  > 
>  > If we can't split things up that way, how about we change one of the two
>  > names? Either make the computer be ofpowerpc or ofppc, since it is a port
>  > to full-fledged Open Firmware PowerPC machines and not the pseudo-OF
>  > machines we buy from Apple, or make the cpu-specific stuff arch/ppc?

> I think...

> arch/powerpc or cpu/powerpc - CPU specific directory

Yes something like this makes sense.
The PPC ISA (the green book) defines the processor behavour down to the
OS level.

> arch/ofppc	OpenFirmware machine (and PSIM)
> arch/powermac	PowerMac
> arch/bebox	BeBox
> (arch/prep	no OpenFirmware machine)

What differences are there?

	o	boot process - ok

		In an ideal world the H/W dependant
		pre-boot would hide all the details.
		Somewhere for each preboot program
		is needed though.
		The thought of the prep preboot loader
		exporting an OF interface comes to mind.
		This is what Sun did!

	o	interrupt controller

		Shouldn't that be a configuration issue?
		[I suspect it is more than that, however
		I did mention that I was going out on
		a philosophical limb :-) ]

	o	HW bus architecture

		Again a config problem?

Have fun with this one :-)

		Andrew

PS: Don't take this too seriously.