Re: LILO

[Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de>]

On Fri, May 11, 2001 at 06:39:58AM +1000, Daniel Stone wrote:
> Um, seeing as GRUB was written because the HURD people got pissed off at the
> lack of a decent multiboot bootloader, I think it would support HURD just
> fine. If there isn't any HURD version, I'd call that a build(d) failure, not
> a GRUB failure.

No, the GRUB and the multiboot standard were developed in parallel by the
same person (Erich Boleyn).  He surely had the Hurd in mind, but GRUB is not
limited to the Hurd.  The motivation was to provide a good bootloader for
Mach and other kernel projects.

One can only wonder why BSD and Linux don't support the multiboot standard.

BTW, as far as I know, GRUB is designed with portability in mind. 
Substantial parts (like filesystems) are not hardware dependant.

On Fri, May 11, 2001 at 06:43:47AM +1000, Daniel Stone wrote:
> On Thu, May 10, 2001 at 04:09:10PM +0100, Frederico Mu?oz wrote:
> > Ummm... ina all this I was trying to found a reason for LILO to be the 
> > default bootloader... every reason
> > I could see appears to be invalid, but there is probably a big one out 
> > there that I'm missing utterly...
> 
> Tradition, and the fact lilo will boot over software RAID, SCSI drives with
> no BIOS, etc?

Dubious.  I have never seen a bootloader for x86 that can boot from a device
without BIOS support.

GRUB development had stalled for a couple of years, but is now active for
over a year again, and it has catched up on all lines.  There really is no
reason to not use GRUB.

> (OTOH, GRUB *does* netboot and multiboot). The reason is this:
> 	* GRUB provides few drivers

Bootloaders don't provide drivers.

> 	* GRUB reads the config file and the kernel off the partition, at
> 	  boot time

Yes, that's one of the great features.  Yuo *can* specify a block list, like
LILO does, so if you really dare, you can load a kernel directory from an
arbitrary store.  Why you'd want to do that when there is support for fat,
ext2, ufs, reiser and minix escapes me. GRUB is also independant of the
drive geometry translation.

It also supports gz and bzip2 (auto detected).

> So, if GRUB doesn't have a driver for <foo> BIOS-less SCSI card (and it
> won't), you won't be able to boot off it. Ditto software RAID, you can't
> boot from that, because GRUB doesn't support md/LVM support, if you want
> that you'll have to use LILO, because it writes the entire boot sector, etc,
> from inside Linux.

I am not savvy on all the glorious details, but I can't find anything in the
LILO doc about BIOS-less SCSI cards.  About RAID: Ditto, I can't find
any special support for RAID disks.  May be very well that GRUB lacks support
for RAID filesystems (so you'd have to keep the kernel on a plain
filesystem), but it certain;y has support for block lists, so if LILO boots
from RAID fs by blocklist (what I'd suspect from what I know about it), so can
GRUB (there is probably no automatic way to do that, but it would be simple
to add).

> Personally, I prefer GRUB, but I can see why LILO is used in some
> circumstances.

Well, any change is painful, as you need to learn something new, and
probably have to fix some bugs after doing the change.  So one good reason
is to not do changes when they are not worth the cost.  But for GRUB, I
think it is well worth it (and not only because I do Hurd).

> PS, the documentation (or lack thereof) for GRUB was absolutely awful when I
> last checked, has it gotten any better?

Even better, there are some new commands which make using GRUB so simple.
You just do "setup (hd0)". 

Marcus

-- 
`Rhubarb is no Egyptian god.' Debian http://www.debian.org brinkmd@debian.org
Marcus Brinkmann              GNU    http://www.gnu.org    marcus@gnu.org
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de