FC4t2 no good without LILO

[Paul Iadonisi]

On Fri, 2005-04-15 at 09:57 -0400, William Hooper wrote:
> Res said:

[snip]

> >
> > its called choice i choose to roll my own kernels i choose to install lilo
> > because grub cant handle some things and cant handle others as well.
> 
> And not having LILO in core limits your choice how?  You are still able to
> download and install LILO, just like you download and install kernels.

  I've been advocating the onward march towards a single bootloader
(that being grub) as well, but I do understand the complaint that it's
too early to eliminate it completely.
  The problem is that the custom kernel option is not analogous to an
alternative bootloader.  I presume that many people who build their own
kernels can boot the original and install the new custom kernel (by
whatever means -- 'make zImage;make modules;make modules_install;make
install' or via a custom rpm).
  Yes, there are some cases where the kernel just won't boot, but the
kernel is such a complex beast that I'm sure many who use custom kernels
do it more out of preference to add a specific patch and/or driver that
isn't required just to get the kernel to boot, but is needed from some
specific, peculiar device.
  The bootloader, on the other hand, is a bit difficult to replace if
it's not, at a minimum, included on the install media.  I do network
installs almost exclusively, so it's not a big issue for me (as I can
just drop the rpm in the install tree), but it can't be assumed that
most people do that (anyone have actual statistics on that, please
present them).
  What I propose is, since it appears the most serious bug (at least in
IMO) -- grub not working on software RAID 1 -- was JUST fixed, that this
new version needs more exposure before eliminating lilo completely.  But
I do strongly suggest that lilo support should be completely removed
from the installer (if it hasn't been already).  I'd even half seriously
suggest that the lilo package should be renamed to something obnoxious
like 'compat-stop-using-this-fricken-bootloader-because-its-going-away-
SOON-lilo-21.4.4-26.i386.rpm' ;-).  And there should be a bright, red,
flashing warning in the release notes that if you are still using lilo,
you NEED to try out grub, and if you have problems with it, REPORT them
and WORK WITH developers.  I.e.: don't call them morons and the justify
it with weasel wording sophistry.  Present real data, not general
statements like 'not reliable' -- that ain't data.
  The ONLY vestige of lilo support in FC4 should be the obnoxiously
renamed rpm and in-your-face warnings in the release notes about trying
grub NOW or forever holding your peace.  And not just trying it, having
it fail, and then just giving up.
  Let me give an example of what your expectations should be, especially
as a tester.  In this current testing cycle, I've come across what
appears to be a regression in the i810 X.org driver.  I haven't got
around to reporting it.  One of the problems is that, especially with X
bugs, it can be (but isn't always) a bit labor intensive for the tester
(as well as the developer, of course).  Unfortunately, I just haven't
had the time to spend on it just yet, and may not have time before
release.  For that reason, do you think I'm going to complain if FC4 is
released with this bug?  NOT A CHANCE.
  This (or a similar) bug existed in FC3 and caused a problem for my
desktop at work.  For that test cycle I *was* helping with the reporting
and commenting on the bug and on testing the fix.  It didn't get fixed
until a subsequent update to X.org.  Was I a little disappointed?  Sure,
but I was patient because I saw the work in progress, largely due to my
own and other tester involvement.  If I, or other testers hadn't
involved ourselves (all the while being RESPECTFUL of the developers), I
doubt there would have been any progress on fixing the bug.
  So for those of you that want what I'm suggesting, but plan on using
the obnoxiously renamed lilo rpm and never looking back ... well, then,
don't expect to be able to use FC5 on your hardware that won't work with
grub today.  For Fedora Core as a project, the world will not end for us
if those machines don't run Fedora Core.

-- 
-Paul Iadonisi
 Senior System Administrator
 Red Hat Certified Engineer / Local Linux Lobbyist
 Ever see a penguin fly?  --  Try Linux.
 GPL all the way: Sell services, don't lease secrets