On Fri, Feb 16, 2001 at 09:26:34PM +0100, Reinoud Zandijk wrote:

 > > 	arch/archidemes
 > > 	arch/acornriscpc
 > > 	arch/cats
 > > 	arch/dnard
 > > 	arch/hpcarm
 > > 	arch/rc7500
 > 
 > Wow .... hmm... for one thing the archimedes, acorn riscpc and the rc7500
 > and the coming imago and omina can easily share code;
 > 
 > The cats, dnard, ofw and ebsa are really more a footbridge/squoia or
 > equivalent and use all pheriphals from isa/pci (AFAIK) and are thus very
 > thin to be a seperate port.

dnard == ofw + sequoia + sa110 + quirks.  It has a way of booting that is
incompatible with the CATS, for example.

CATS == Cyclone FW + 21285 + sa110.  It has a way of booting that is
imcompatible with other 21285 + sa110 systems running non-Cyclone
firmware.  A 21285 + sa110 that is NOT using Cyclone should be a
separate port.

Note these systems also have different disklabel requirements, etc.

There is no reason to force these things into a single port if you
can't make a single GENERIC kernel/boot loader that boots on all of them.

You can still do code sharing, but it's more intuitive to the users
if you cleanly separate things (and have correct install notes for
each type).

Now, perhaps my rc7500 example was a bad one -- if it's a system that
also runs RiscOS and boots just like a RiscPC, then perhaps it's
reasonable for it to be in an "acorn32" port (and the archidimes in
an "acorn26" port?)  ..  but I was always under the impression that
the rc7500 was a funky eval board (and, it's clearly incompatible with
the RiscPC, since there is a separate kernel for it).

Sure, it can share e.g. IOMD code ... but that can be done via
sys/dev/acorn/iomd/... or something.  See how the sparc and sparc64
ports share Sbus device drivers (sys/dev/sbus).

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>