[3513] in SIPB_Linux_Development
Some concerns with the SIPB redhat installer
daemon@ATHENA.MIT.EDU (John Hawkinson)
Sun Jun 17 23:31:33 2001
Date: Sun, 17 Jun 2001 23:31:30 -0400 (EDT)
Message-Id: <200106180331.XAA25875@multics.mit.edu>
To: linux-dev@MIT.EDU
From: John Hawkinson <jhawk@MIT.EDU>
Hi, folks on the "other side",
I've just watched (still watching) a user struggle with installer
linux using the SIPB installer and take up a few hours of phone time
in the SIPB office, and cause a lot of agony for Angie, alex_c, and
myself, as phone gateways.
This particular user is not the most clueful of linux users, and
has encountered a phenomenal level of difficulty, but it has demonstrated
some problems in the install process. I bet Angie and Alex can comment
on more, but I see two significant ones:
#1: The SIPB installer allows you to install on a machine where the
linux partition is past 8gb into the disk, and lilo falls apart
completely and leaves you with an unusable MBR (prints "LI"). To my
mind, this is completely unacceptable. Is it not possible to just hack
the installer to verify that the linux partition is early enough and
punt otherwise, so you don't leave the user with a disk that won't
boot usefully? Or at least emplace prominent documentation notes about
this? It really sucks to be in this situation, and it sure seems easy
to do.
#2: This user had a pre-existing Windows partition and wished to
install Linux in the remaining free space in her hard drive. She had
_great difficulty_ interpretting the installer instructions on how to
set up the partitioning. I think that much of the fault lies in her
lack of understanding of them, but some of it falls in the instructions
themselves, and the requirement for them:
a) The help screen talks about FAT partitions and "the root" and "/"
and it is not sufficiently clear that "/" is "the root" and that you
must select {whatever} and label your linux partition /. Nor is it
sufficiently clear that a swap partition must be created and of what
size.
b) It seems relatively broken to me that the user has to go through
manual partitioning in this case anyway. It seems like a good
enhancement would be to automate this (hey, the NetBSD install had
this; I need to do more work on that. competition, *nudge* *nudge*).
c) I forget what c) is.
Again, I think that #1 really needs to be dealt with in the short term,
and that #2 is a much more "would be nice" thing.
Thanks.
I apologize if this is misdirected, btw; please let me know where else
I perhaps should have sent it.
--jhawk
