Soda forwarded your mail to me.

> I recently aquired (for free) a MIPS R3000 motherboard, with on-board
> SCSI (AIX6150), Lance Ethernet, floppy, 2ser, 1par, a 96pin connector
> for graphics (and the graphics board).
>
> I wonder whether this is an ancestor of the ARC design...

I think it's not an ARC animal.  As Soda pointed out, it seems a decendent
of SUMIstation family, one of oldest MIPS based UNIX WS in Japan.

> it was built
> by Sumitomo Electric, and integrated into a workstation by ITOS GmbH
> (who was member of ACE, I think) in Germany, who no longer support
> them and even no longer have docs archived.
>
> The disklabel format is, however, a RiscOS one.

It might be a straight port of MIPS Co. RiscOS.  It means, the machine
is Big Endian unlike NetBSD/{pmax, pica} or OpenBSD/arc.

> Do you have any information about these machines? I'd like to run
> NetBSD on it eventually (although I have a backup of the native OS,
> and an install floppy with sash on it).

It's nice to have NetBSD/big-endian-mips.  Endian related issues were
once solved when one of our common ex-colleague Kazumasa Utashiro
ported dying CSRG 4.4BSD to SONY R3000 based UNIX WS.  You can reach
his effort grabbing 4.4BSD-Lite2 distribution and looking at news3400.

You might notice the 4.4BSD for SONY NEWS is imcomplete and so defunct now.

      * the port was done using device driver source code donated from
	SONY real product line. Unfortunately, original source code quality
	is not so good as todays NetBSD.  The news3400 port was suffered
	from technical/business shortsight in real world.  Besides endian
	issues,	it's not worthwhile to look at news3400 4.4BSD Lite source
	code.

      * NetBSD/mips seems toward endian free, but currently code depends
	little endian.  Your efforts would contribute general NetBSD/mips
	quality.

As a side note, I've once ran a personal project to port
4.4BSD-whatsoever-late to MIPS Co. UNIX WS.  It was abandoned mainly
because my lack of experience and knowledge, however I learned much about
withdrawn MIPS Co. product line.

I guess your project could follow steps like below.

- Learn hard the machine first.  If the machine is RiscOS, it must contain
  device driver source codes in its filesystem.  Grab it. They would be
  great help to your work.

- Build NetBSD toolchain on hosting machine.  NetBSD/pmax would be best,
  but cross compiling on other NetBSD platform is fine.  Don't forget
  big-endianness.

- Plan conf/files.XXX and conf/GENERIC file carefully.  At first most of
  files would have just stubs.  Build kernel with minimal code so as to
  boot and check several trival things (then die itself); measure memory
  capacity, CPU types, or whatsoever you guess.

- If the Sumitomo WS can network boot, you have a very good luck.  Netboot
  is first, then track with SCSI issue.  MIPS Co. box had complicated
  sash (stand alone shell) including network boot capability.

Just now I remember there is a Sumitomo bland UNIX WS gathering dusts in
our computer equipment storage room.   I'm going to see what the machine
reallly is, and wether I can help you more.

Tohru Nishimura
Nara Institute of Science and Technology