This may not be so much booting the system, as learning how to set it 
up appropriately in the first place (or at least learning to use 
OpenBoot on something that isn't a Sun for a change)...

My system looks like this (it's a 500MHz Sawtooth G4 AGP, currently 
512MiB RAM):
wd0 = 28.0GB Seagate ST330621A #this is the NetBSD target
wd1 = 19.1GB Maxtor 2B020H1 #single HFS+ slice
sd0 = 17.1GB Quantum Atlas 10K 18WLS #two slices; HFS+ for Mac OSX and 
an OpenStep-style UFS slice

So on the Seagate I used OS X disk utility to create a 4.0MB HFS+ 
slice, upon which I copied /path/to/cdrom/ofwboot.xcf and the 
netbsd-GENERIC.gz kernel from my install CD.  The rest of the disk was 
marked (from Darwin's POV) as unused space.  Boot from install CD, 
select vt100, select (I), choose wd0 as the disk, offered some 
partition maps for "the NetBSD part of the disk", and chose 32MB /, 
2048MB swap, rest=/usr.  newfs goes fine.  Select to install all the 
tgzs from CD-ROM, this works, get to the reboot and (unsurprisingly) OS 
X pops back up.  Only my 4.0MB HFS+ slice seems to have 
disappeared...Darwin is seeing a single slice (disk0s2) which consumes 
the entire disk.  Presumably I made a wrong turn at the partitioning 
stage...my experience of installing BSDs (apart from Darwin) is on x86 
and my understanding there is that the BSD will create itself a primary 
partition, then generate BSD partitions *within* that.  Is this not 
what happens on the Mac?

Is my NetBSD-macppc system now bootable through some OF invocation; 
perhaps with the ofwboot on the CD-ROM?  If not, what should I do 
differently through the installation stage to end up with a bootable 
NetBSD at the end next time?

Cheers,

Graham.
-- 
Graham Lee
UNIX Systems Manager,
Oxford Physics Practical Course
http://nextstep.sdf-eu.org