Re: GRUB for Debian in general?

[Jeff Sheinberg <jeffsh@erols.com>]

Hello Joel,

    I am a great fan of grub, however, I would like to correct
some misinformation you have been spreading regarding grub.

Joel Klecker writes:
 > As I said, the current grubinst works only with grub 0.4, and grub 
 > 0.5 is a vast improvement on 0.4 (among other things it brings grub 
 > up to full feature parity with lilo with regard to Linux).
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    1. Grub cannot switch hard drives in the BIOS like lilo does.
For example, one cannot boot DOS from the 2nd hard drive, fooling
DOS into thinking that it was booted from the 1st hard drive.
Lilo does this blindfolded and with one hand tied behind its back.

    2. Grub does not (yet) recognize Linux extended (0x85)
partitions.

 > GRUB's partition naming makes sense if you get used to it, all it 
 > gets from the BIOS is the disk number, so it has no idea what kind of 
 > hard disk it's looking at, so presenting them all as hdX makes sense 
 > (it gets really perverse when you note that the order in which the 
                   ^^^^^^^^
 > BIOS presents the disks depends on the configuration of the BIOS).
 > 
 > And as for partitions, there is no way around the fact that BSD 
 > partitions are different than standard PC partitions, GRUB needs to 
 > differentiate them somehow. Again, there is some perversity in that a 
                                                    ^^^^^^^^^^
 > set of BSD partitions can be inside a single PC partition.

    1. Grub's partition naming scheme is zero based, going against
established convention.  This has nothing to do with BSD
partitions, which are just another form of extended partition,
just like DOS extended (0x5), and Linux extended (0x85).  DOS and
Linux just access these logical partitions by starting counting at
5, 6, etc, while BSD starts counting at a, b, etc.  I agree that
the BIOS booting procedure may be a little perverse, but I don't
see where this affects grub's partition naming scheme.

    2. Grub's partition naming scheme, being zero based, is fine
for some people, it just causes unnecessary confusion, because
there is not yet any way to refer to partitions by their
conventional nomenclature.  In addition, grub's naming of hard
drives is also zero based, likewise going against established
conventions and causing a lesser amount of confusion.

    3. I propose the following solution, just add a new naming
scheme to grub that adheres to established conventions.  I call
this new name `hdd'.  Thus, (hd0,2) == (hdd1,3) .  This simple
addition of the `hdd' name to grub would be backwards compatible
with what we have now, and allows one to chose the scheme that she
finds the least/most confusing!

    I would also like to point out the area where grub is more
flexible than lilo - grub has much less reliance on statically
calculated data.  Grub locates partitions, kernel images, and
files at run time, whereas lilo is constrained to do the same at
map creation time.

Cheers,

-- 
Jeff Sheinberg  <jeffsh@erols.com>