> So instead of probing the devices itself or asking the BIOS or asking
> OpenFirmware, the kernel has to ask I2O (or something else), yes?

Yes, but I2O is presumably documented, so you don't need the BIOS as much.

> PCI is pretty easy to re-probe if you're not looking to allocate BARs or
> load FCode ROMs.

Works for me. I guess the real problem is getting this damn ISA-PnP
situation under control, then. That, and putting PCI bridges into
new PC's so you never need any ISA slots in the first place.

> The way things are going, it may well be easier to simply burn *BSD into
> the ROM and let it be the firmware as well just to avoid replicating
> code. :-)

Wow, I didn't have to be the one to suggest this!

There's already enough flash in most machines to hold a trimmed install
floppy image.

I would be a lot happier if the PC BIOS could be replaced with a "TINY"
kernel booting from '/rom'. After scoping out the system it would pause in
a full-featured boot manager; if I hit SIGINT then it drops into a shell
and I can run format, disklabel, ifconfig, telnet, tip, sysinst, etc.

Hell, why not just make a 1.44 meg partition on your boot drive, put a
hacked up copy of the install floppy there, and use that as your boot
manager?

I guess the only real problem is that bootup still takes a _lot_ of time.
What takes so long in the device prober anyway? It is timeout loops for
every device that isn't detected? Why _does_ the 1.3.2 atapibus0 dmesg
pause for many seconds before continuing with the boot? Why? Why? :)

Todd Whitesel
toddpw @ best.com