On Sat, 25 Sep 1999, Frank van der Linden wrote:

> On Sat, Sep 25, 1999 at 08:44:56AM -0700, Eduardo E. Horvath wrote:
> > I don't like the idea of moving the partition reading code into the
> > bootloader.  
> 
> That code won't be in the bootloader. All that the bootloader would know
> is some <device, start, end> combination, and pass it on to the kernel.
> These values would be determined by installboot, for example, as is done
> now on the i386 port to write information into the first stage on
> where it can find the 2nd stage.

In the sparc64 port, neither the bootblock nor the bootloader has this
information embedded in it.  Both interpret both the native and the NetBSD
disk labels to determine whiche file to load.  They were done this way so
moving a few files wouldn't require reinstalling the bootblock.

> Yes, you would lose RB_ASKNAME functionality this way, this was already
> mentioned in the original discussion. If many people rely on this
> functionality (I'm not sure many people do, besides some developers),
> we would need to include optional default wedge construction support
> for some partitioning types (the "native" ones for the port) in
> the kernel.

I really don't want to give this up.  If I can't share partitions between
NetBSD and Solaris I might as well switch to Linux.

=========================================================================
Eduardo Horvath				eeh@one-o.com
	"I need to find a pithy new quote." -- me