on,  Thu, 13 Nov 1997 20:51:09 -0500 (EST)
  maximum entropy <entropy@zippy.bernstein.com> writes:

Date: Thu, 13 Nov 1997 20:51:09 -0500 (EST)

>I untarred the sysinst.tar.gz on a fileserver and netbooted a 2100
>from that, then dd'd the diskimage to a local disk.  I then booted
>from that disk and sysinsted.  Here's some feedback on my results.
>
>- It isn't possible to sysinst to the disk you just installed from, so
>this method requires two disks.  I realize it's not reasonable to
>expect this to work, but it isn't documented in the pmax install
>NOTES.

If I'm understanding you, this *does* work, in several  ways:

	 1.  If you've netbooted from the diskimage.tar.gz, you should
	     also be able to run sysinst from the diskless root,
	     and label the local disk.

If that's not clear, it should be documented better in the NOTES and
in a real install document.


	 2.  If you  dd the raw diskimage onto the raw partition
	     of a disk and boot it, you can run sysinst from the
	     `a` partition of that disk -- with the current root
	     the same as the installation  target disks.
	     This should Just Work.

	     (the alternatives for single-disk systems that
	      can't netbot are hairy: dd the diskimage
	     onto the disk twice, once to the raw disk to set 
	     disklabel and bootblock, and a second time to dd it
     	     into the swap partition, 65536 blocks from the beginning
	     of the disk.)

If that's not clear, it should be documented better in the NOTES and
in a real install document.



>- Before starting sysinst you need to remount root read-write.  This
>should either be documented or sysinst should do it for you.  If root
>is left readonly, sysinst will fail to mount the partitions for the
>new disk after newfsing them.

I guess so. Perhaps sysinst should just complain and exit if the root
is not mounted read-write. read-only NFS roots are sort of plausible,
though.

I haven't looked at the other points yet; thanks very much for the
feedback, though.