Bill Studenmund wrote:

>The thread was focusing on what to do about the overlap between a struct
>disklabel in sector 0 and the driver entries. For new disks, we are not
>required to continue puting a struct disklabel on there, so we can avoid
>the problem. :-)
>
Ah!  My misunderstanding then.

>In general, I'd like to depreciate shoving in struct disklabels in favor
>of using Apple Partition Maps. We should always be able to READ struct
>disklabels, but I think our default should be something more compatible
>with the other OSes which run on our hardware. :-0
>
We're on the same page then and I have no objections at all.

>Hmmm... There is the difficulty that bootxx, the loader which we shove
>into the partition, won't work without open firmware. :-( The StartManager
>booter is quite a different beast than the OF boot loader.... Hmmm....
>
Well if the Start Manager can be used, what's to keep us from having 
installboot adjust Block0 and the driver table to get the Kernel loaded 
from the selected root partition? If we still need a secondary loader 
couldn't it be stored in a spearate Partition named something like 
"NetBSD_Driver"?

>
>If we have a struct disklabel in block 0, we can't have more than one
>"driver" listed in the driver table. The second entry's length will
>overlap with the disklabel magic number.
>
Is the driver count in Block0 being updated so the Start Manager doesn't 
look at the additional entries?  Not that this would keep NetBSD from 
booting, but it might screw up MacOS from booting at some point.

-bob