Bill Arbaugh <[EMAIL PROTECTED]> writes:
> >
> > A possible alternative would be to use some of the code from oskit.
> > The only downside is last time I looked it didn't looked like it
> > wasn't noticeably smaller than just using a minimal linux kernel...
> >
> Jay Lepreau and I have been having some discussions about this very
> topic. Essentially, OSkit will be about the same size as minimalized
> kernel.
>
> > What do you see as the gain of using GRUB versus a small
> > user space program, on top of a minimal kernel?
> >
> None, and your suggestion is exactly the way we're moving.
O.k. If the goal is booting over the network, I currently have a
netboot package that does dhcp and tftp. And downloads a kernel over
the kernel and boots it. The ramdisk compressed was only 10k, last
time I looked.
For other kinds of booting control it should be relatively simple to
build on top of a linux booting linux infrastructure.
I can currently build a minimal 2.4.x kernel which compressed comes to
about 360K. If I could get small enough to fit on a 256K rom I would
love it but unfortunantely I can't do that today.
What kind of sizes are you seeing?
In general I'm interested in this area, and I'm really pushing to
build a solid bottom layer on which all of these kinds of things can
be built. If someone can build a better bootloader on top of the
infrastructure I'm building, I'd love to see it (especially because
you become multiple architecture almost effortlessly). Until then
I'll just build the pieces as I go.
Eric
