Well, I've gone and shot myself in the foot again. ISTR reading
something about a dual-boot system, or at least two different
kernels, and someone said that they have to have different names.

OK, fair enough.

But, I thought, isn't having them on different partitions enough?

i.e.: Here's the relevant parts of my lilo.conf:

# Start LILO global section
boot = /dev/hda
prompt
timeout = 50
vga = ext
image = /boot/vmlinuz
root = /dev/hdb1
label = Slack-10.0
read-only # Non-UMSDOS filesystems should be mounted read-only for checking
other = /dev/hda1
label = W2K
table = /dev/hda
image = /boot/vmlinuz
root = /dev/hdb8
label = Slack-9.1
read-only # Non-UMSDOS filesystems should be mounted read-only for checking
image = /boot/vmlinuz
root = /dev/hdb7
label = Boot0
read-only # Non-UMSDOS filesystems should be mounted read-only for checking
-------------------------

I guess I had assumed that when I pick option 1, it loads vmlinuz from
/dev/hdb1/boot, option 3, /dev/hdb8/boot, option 4, /dev/hdb7/boot.

Obviously, I was wrong, because now, (this lilo.conf was done while in
10.0, and written to the MBR) when I select option 3, boot Slack 9.1,
it loads kernel 2.4.26, which is on /hda1.

Does that mean that lilo itself reads the "image" values before it
decides which partition to boot from? Like, it can only get it from
hdb1?

So, I should put:

image = /boot/vmlinuz-2.4.22
root = /dev/hdb8
label = Slack-9.1
read-only # Non-UMSDOS filesystems should be mounted read-only for checking
image = /boot/vmlinuz-2.4.22
root = /dev/hdb7
label = Boot0
read-only # Non-UMSDOS filesystems should be mounted read-only for checking

in my lilo.conf, and copy the 2.4.22 kernel over to /dev/hdb1 with that
other name? It sounds logical, and also the way to try other kernels with
the same rest of the system, but somehow my misunderstanding of the way
lilo works feels like a gotcha.

Thanks,
Rich