FC4t2 no good without LILO

[Mike Bird]

On Wed, 2005-04-13 at 05:28, Jeff Spaleta wrote:
> If you have other SPECIFIC issues with the currently aviable grub
> package in rawhide, raise them and they will be evaluated case-by-case.

In the real world, a history of unreliability is a SPECIFIC issue.  It's
the SPECIFIC issue that gets Windows servers replaced by Linux servers.

Peter in his cubicle may believe that Grub is now perfect.  Even if he's
right it's irrelevant at this time because nobody (except unwitting
Redhat shareholders) is going to bet the farm on a new version of a
product with a history of unreliability and missing features.

A year from now, if Peter's dreams are realized, would be the time to
consider removing Lilo.  Or maybe that would be the time to start
replacing Grub Legacy with Grub2.

FC4t2 Grub includes a Dec'04 MD patch, a Jan'05 MD patch, a Feb'05 MD
patch, and a Mar'05 MD patch.  We all hope that no more patches will be
needed for MD in Grub, but is it realistic to bet the Redhat farm on it
today?

I'm certainly not going to bet our clients' data on Grub until I see how
reliable Grub is now and also whether the next few updates are reliable
too.

A HISTORY of reliability is essential for any mission critical package. 
A proven history over multiple upgrades is particularly important for
Grub which uses a fragile shell/awk/sed script to parse mdadm output
instead of copying the simple GET_ARRAY_INFO ioctl used by Lilo.

The appropriate course is for Peter to restore Lilo, apologize on behalf
of Redhat for taking five years to add MD support to Grub, state that
Grub is now believed to support MD under such-and-such SPECIFIC
conditions, and ask us to give Grub another try.

--Mike Bird