I put today's kernel up for anonymous FTP, in
ftp.dsg.stanford.edu:pub/NetBSD/arch/pmax/snapshot/netbsd.

Be warned: this kernel is built from my own tree, and includes NTP
kernel support, which works very solidly for me on a 5k/240.
It's likely to be slower than previous kernels on a 3100;  the
time-of-day interrupt rate went from 64Hz to 256Hz.

I'm really not sure what the "colour problem" is.
It may be that the 3100 pm driver and the cfb driver
use incompatible ways of interpreting a colour.   But if this
is in glass-tty mode, I have no *idea*  what the problem is.

I hope to put out a precompiled kernel with a fix for the pmax
lance bug tomorrow.

--Jonathan