On Wed, Dec 29, 1999 at 10:56:21AM +1100, Simon Burge wrote:
> Bill Studenmund wrote:
> 
> > On Wed, 29 Dec 1999, Simon Burge wrote:
> > 
> > > > For an install kernel (a kernel using a memory disk root filesystem),
> > > > you're right. I think there's been some interest in it. Like some variant
> > > > of tar which would use an ffs as the output format.
> > > 
> > > Curious...  I don't recall anything recently on any NetBSD lists - do
> > > you have a pointer?
> > 
> > Mostly just mutterings with other developers. Maybe on icb or in personal
> > communications. I don't think anyone has any code yet. :-)
> 
> Hmm, must get onto icb every now and then...
> 
> > [ installboot ]
> > 
> > The versions I'm familiar with use both the mounted file system and its
> > raw device to figure out the blocks backing the file. Great if it doesn't
> > need to. :-)
> 
> It makes sense now (the coffee has kicked in)!  We're talking about
> first stage boots along the lines of the i386, where the block numbers
> are hard coded.  My answer would be to use a different architecture, but
> I'm sure that won't win many votes :-)

Mmmm, different architectures.

Unfortunately, with them you can have 

- affordable machines
- fast machines
- affordable power consumption

pick any two.
	-is