Perry suggested that I repost this message to tech-install.

I should have posted here first, but I wasn't thinking at the
time... ;-)

Cheers, 
Dean.

------- Forwarded Message

Dean Huxley <dean@huxley.org> writes:

> My brother-in-law and I just installed NetBSD on his machine, and I
> made some notes as I went along.  We used the boot disks from the
> latest snapshot-19990315.
> 
> Unfortunately, our first attempt newfs'd his NT partition.  Strangely,
> he didn't seem too phased by that ;-)
> 
> For the most part we just accepted the defaults that sysinst gave us.
> We chose that we wanted to install on part of the disk (instead of all
> of it) and the default disklabel that it gave us apparently was for a
> full disk install.  We had originally selected that we wanted
> cylinders reported and somehow it switched back to using megabytes.
> We were a little confused by the numbers being different than we
> expected but we assumed (incorrectly) that sysinst knew what it was
> doing.  We proceeded to do the install, and quickly destroyed the NT
> partition (it happened to be the first one on the disk).
> 
> We verified that his other three partitions were still untouched,
> including the one where we actually wanted to install NetBSD.
> 
> On our second attempt, we inspected everything that sysinst told us.
> That was when we discovered the megabyte/cylinder default switching.
> Since we had already blown away the first partition (NT) we decided to
> install on the first partition this time.  We chose the "megabyte"
> disklabel reporting option and started seeing some consistent numbers.
> Again we choose a partial disk install.  When the default disklabel
> came up, we finally understood it was still trying to make a "complete
> disk" disklabel.  We shrunk the size of the /usr partition to fit
> within the bounds of the old NT partition, and then the install ran
> much better!
> 
> The only other complaint I have, is that the ftp method could be made
> easier to use.  First, the default directory didn't point to the
> snapshot binary sets, but I guess that might normal when it's not a
> formal release.  Having to hit a key after each successful ftp
> transfer was a pain.  Unfortunately, on one of the final X tar files,
> the ftp server we were using seized up.  I had to abort the transfer.
> I went back into sysinst to switch to a different ftp host and try to
> finish ftping, but it failed the mount tests because /mnt and /mnt/usr
> were already mounted.  I unmounted them and tried again.  It passed
> the mount tests, but then proceeded to download ALL of the tar files
> from scratch!  I had hoped that it would resume where it left off...
> 
> Eventually, I managed to get all the tar files and untar them.
> 
> I got the message that I'd have to edit rc.conf to change
> rc_configured=NO.  Does a normal successful install do the same or does
> it actually setup rc.conf for you?  (I've only done a couple installs
> lately and never had a completely successful, uninterupted install, so
> I'm not sure what is supposed to happen...)
> 
> All this being said, I still think that this sysinst is much better
> than the last one I used (early December sometime).  
> 
> I'm not complaining - I'm giving feedback ;-)
> 
> Cheers,
> Dean.

------- End of Forwarded Message