> Warning:  venting ahead.
 > 
 > People (not me) post on the netbsd newsgroups asking for help installing
 > netbsd-pmax.  Response:
 > 
 > "Send mail to the mailing list."

Well, that's good advice at least; speaking only for myself there's
a big chance that some articles on the NetBSD newsgroups will expire
before I ever get to read them, let alone answer to them.  If you
ask for help on the mailing list, I can at least go back later and
send a follow-up mail.

 > People send mail to the mailing list, and are ignored. 

This is certainly bad, and I feel a bit guilty here.  Many of the
people who've had problems has had very equal-sounding problems
(with the initial install/bootblocks/disklabel stuff).  I made
a set of installation instructions around newyear, and then I
promised I would try to reproduce the problems people reported.
However, in my new job as Senior Engineer at the math. department
there are lots to do, and no DECstations.  So with my small sporadic
efforts I never managed to reproduce the problem people had.

In retrospect, it would have been better to answer those people with
"Others have the same problem, but we don't know why yet.  Subscribe to
the mailing list and be patient".  Since I (mistakenly) always felt an
actual solution was just around the corner, I didn't send those letters.

 > I -know- you're volunteers and are not here for our convenience.
 > I appreciate the work you're doing.  But what would be nice is if
 > no one -knows- how to install NetBSD Pmax, and no one -knows- why
 > Arne's latest instructions fail (by the way, thanks Arne -- I 
 > do appreciate your giving it a shot!), then -SAY THAT- so that those
 > of us who have been considering netbsd-pmax for use in a production
 > environment can quietly cross it off of our list and go on our way.

Well, if you need NetBSD/pmax in a production environment, I think
you'll either have to put in some work yourself or wait a bit yet.
However, hopefully the initial install problems will be solved shortly.

 > Those of ous who are considering NetBSD-pmax aren't, generally,
 > novices.  We don't need a tcl/tk/X based installer which guides
 > us gently through fields of flowers.  We're NOT AFRAID of
 > disklabeling, newfsing, or otherwise getting our hands dirty.

Good :-)

 > But having an installation which is at least POSSIBLE from an
 > Ultrix environment is as important to the NetBSD-pmax team
 > as it is to the potential users.  If a user can't install
 > NetBSD-pmax, all the user loses is an extremely optional
 > OS.  But the NetBSD-pmax team loses more -- that is, potentially valuable
 > contributors of time and energy.

I agree.  What NetBSD/pmax needs most (more than installation tools,
more than serial consoles, more than X11R6 server with support for all
possible graphics cards) is some more people working on it.  I'm afraid
I'll only be able to work on it in my holidays (Easter, some weekend
in May, maybe some more weekends in July, like), and Jonathan can't do
everything alone. (Though he valiantly tries - thanks again for all that
you do, Jonathan!)

Well, enough meta-technical discussion, here's the good news:
I've finally done what I should have done a month and a half
ago: I've bought a 1GB disk and put in in a machine that normally
runs Ultrix in a nearby student lab.  Then I tested my installation
instructions on this machine.  Of course, they worked perfectly,
first try.  However, getting some hunches from the various
problem reports, I then started from scratch again, but before
trying the NetBSD install, I used the disk from Ultrix a bit.
(chpt, newfs /dev/rz3c, mount, fill with data, etc).  Then did
the installation checklist again.  Ka-Boom!  So it only worked
the first time because I started with a completely blank disk.
I was therefore missing a step:

-* 0 *-
   Zero out any old Ultrix partitioning information and/or bootblocks
from the disk, by doing:
   dd if=miniroot-1.1+tar+gzip+mount_ffs of=/dev/rrz2c bs=512 count=64
[I'm not sure if you should use count=32 or just dd the whole thing for
sure, but 64 worked for me.]

After this, using `chpt -q /dev/rrz2c' should complain:
"No partition table found in superblock...
 using default table from device driver."
while `disklabel -r rz2' should give just "no disklabel".


I'll resend full installation checklist ala' AHJ again,
hopefully this bit of information will help someone over the
barrier.  Then, if people report that this procedure works
OK, we can make some easier-to-use scripts to automate it.
Alternately, we can try to produce a more complete miniroot that
can be used more directly (as Manuel suggested).  That may be
even better.

  -  Arne H. J.