A rant regarding LILO and various other related issues

[John Morris]

Wow, been watching this flamewar for days now, hope anyone can join.

On Tue, 2005-04-12 at 21:23, Peter Jones wrote:

> You also need to look at all the other packages involved in setting up
> and maintaining the bootloader to.  They're pretty complicated, and a
> lot of work.

I can certainly appreciate that.  I'm certain it complicates life for
the maintainers of a lot of packages to cope with two bootloaders.  But
I also don't understand earlier posts whinging how lilo is
'unscriptable' either.  Both use a flat text file with a header section
and a stanza describing each boot opton.  You do have to invoke lilo
after any change but a scripted solution should never forget a step.

> > Certainly a five year old version of Lilo handles software RAID better than
> > today's Grub.
> 
> This is just plain insulting.  You haven't even *tried* the software
> RAID support in today's grub.  You've already said as much today!

Ok all you brainiacs at RH, could you tell me what dark magic I need to
incant to get "grub-install" to 'just work(tm)' on a RHEL3 system
running software RAID1 on a pair of IDE drives?  So far the only ways I
know is to either let anaconda wizardry deal with it when initially
installing or use the grub shell and a string of rather newbie
unfriendly commands I got from a GRUB+RAID Howto I Googled up.

And here is where a lot of the vitriol is coming from amongst certain
users; we feel that, for reasons beyond our ken, Grub was pushed into
service as the "One True Bootloader" years before it could easily deal
with what for us are everyday tasks.  Once you have been burned a couple
of times by the unreliability of the modern IDE hard drive, RAID1
becomes a routine item.  For us RAID isn't an advanced feature for
servers, it is an essential item that any serious workstation needs
right along with a good UPS.  But for years every time anyone has filed
a bugzilla or complained on a list that grub ain't cutting it out in the
real world they either get ignored or blown off and get the party line
again; that Grub WILL replace LILO, that it really is better if only us
poor rubes could RTFM, and that at any rate we had better just accept
that it is a done deal.


[MODE=RANT kinda off on a tangent]

This is a symptom of a bigger problem in the whole OS/FS world, the
tendency to toss a solution that has some issues but works for a lot of
people for one that doesn't work as well at the moment but is believed
to show better promise.  So we have dozens of media player apps, all
'mostly sorta finished.' and which one will ship in a particular version
of a particular distro is apparently more random chance or a key
developer's fav at the moment than based on anything rational.  We get
dozens of package manager apps, all in various states of disfunction,
etc. etc.

(Oh, and while on the subject of media players... can anyone explain
what feature HelixPlayer provided that ANY of the major Free players
didn't already have? But there shouldn't be two of anything else because
that might confuse end users or consume excess development resources. 
So is Helix a paid product placement by Real or what?  Inquiring minds
wanna know!)

And don't even get me ranting about the everchanging state of graphical
system administration.  Just in the Red Hat linage there have been how
many ways to graphically configure a network interface?  Have ANY of
them actually lived long enough become both feature complete AND
actually work in all situations before being tossed aside for the next
shiny new total rewrite?  Long ago learned that the command line was the
only interface that would be even remotely stable from release to
release so I don't really know the answer.

How many here had angry clients when linuxconf went away for the latest
rewrite of all of the system admin tools and no longer had a web
interface that the proles running Windows could use to add accounts to
their mail/file/etc server?

Or just how many window managers has been the Red Hat default?  Off the
top of my head I think the list is fvwm, fvwm2, fvwm95, enlightenment,
sawmill, sawfish and metacity.  I think metacity probably holds the
record for the number of consecutive releases.... which means it will
probably get replaced soon.

Or a pet peeve of mine: the humble gnome mixer applet.  The version on
my laptop now allows me to relabel the sliders to match reality, for
example I have one slider that controls the internal speaker on my
laptop, another that controls the line out plug with the external
speakers.  The DVD drive in the dock comes in on line2.  Better not
upgrade to the latest one (RHEL4) though, because it appears that it has
been rewritten yet again and doesn't have ANY preferences at the moment.

Bottom line, change is sometimes a good thing.  But change for the sake
of change isn't.  And before removing the existing solution the
replacement should actually be mature enough to replace the old version
in all common and most not so common usages.  Both Fedora and RHEL are
straining the capacity of DVD.  We passed the point where we need MORE
software long ago, we need the existing software to work.

By now Grub has matured to a point it probably makes sense to be talking
about deprecating LILO, but certainly not back in the RH[789] days when
LILO first went on the list of packages scheduled for termination.  Back
then the only thing Grub offered was a VGA splash screen, and THAT wee
bit of fluff was not worth the loss of features in LILO to gain.

[MODE=NORMAL]

-- 
John M.      http://www.beau.org/~jmorris     This post is 100% M$Free!
Geekcode 3.1:GCS C+++ UL++++$ P++ L+++ W++ w--- Y++ b++ 5+++ R tv- e* r