grub and alpha tar

Paul Roberts dagmar at speakeasy.net
Thu Oct 10 16:35:58 PDT 2002


On 8 Oct 2002, Billy O'Connor wrote:

> You're between a rock and a hard place on the grub/lilo issue.
>
> 1. Move to grub -- "I want scary tar!"
>
> 2. Keep lilo    -- "I want new lilo!"

I finally got done dorking around with Grub this week, and here's my take
on the situation...

While grub is technically more flashy, it's got some cons to it that I
don't much care for and can only see as _causing_ more problems than they
would solve...

  - No official support for graphical boot screens (this will become
   somewhat more important over the weekend).

  - It's nowhere _nearly_ as straightforward as LILO to configure.  This
   is almost sure to generate lots of lovely new FAQ questions.

  - It's still technically beta software, and (pardon the pun) to boot
   not as time-tested as LILO.  It would be a fairly bad precedent to set
   to tell people to use "some" beta-release software on production(*)
   equipment unless there is some specific security or stability issue
   that is solved by using it.

The arguments that are in favor of Grub seem pretty minor to me... The
docs for it make a big deal of being able to not having to specify the
amount of memory in a system manually, but let's face it... machines
afflicted with problems seeing beyond 128Mb of RAM are quite rare, and
I've not seen the problem occur with anything of recent manufacture.

Yes, so it can use serial ports like a serial console.  Good for it.
This is of marginal usefulness at best for most people.  It's of almost
zero usefulness to LFS that I can see because people who are going to need
to be doing those sorts of things on a headless machine are definitely few
and far between, and mainly they should be competent enough to read the
copious grub documentation and figure it out, or they probably shouldn't
be messing with such things.  This goes for the network booting aspect of
it as well.  (The odds are that they'll probably already know how to do
this if they need to, and it's way outside LFS scope.)  To make this
completely clear, the only shortcoming that can be attributed to LILO in
this aspect is that you just can't pick any LILO options when booting from
serial (to my knowledge) but both can be used on a machine that has a
console on a serial port.  To make matters worse, a lot of the newer
motherboards are starting to support redirecting the screen to the first
serial port in the *bios*, making the point entirely moot. (**)

...and yes, great, you can pick your own kernel at boot time without
having to add a menu entry for it.  How many people actually need (or
would even _use_) that feature?  Most people would _not_ be using
it--it's likely the vast majority would never need it.

It's for these reasons that I think LFS should stick with LILO alone until
such time as Grub is actually given the 1.0 of approval, and then simply
add Grub as a more complex option for people to have available to them if
they decide they need those few extra features Grub has.





(* There isn't much sense in teaching people anything but a "production"-
  quality approach, since they won't have anything to unlearn later if
  they branch out into doing more risky development work, and it cuts down
  on the number of bizarre things that can go wrong and distract them from
  learning.)

(** The vast majority of the machines I've used Linux on have been
  headless, and not once has a lack of LILO control been a problem.)

-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe lfs-dev' in the subject header of the message



More information about the lfs-dev mailing list