Teaching how to install Linux to ... Linus :-)

Hubert Tonneau (hubert.tonneau@easynet.fr)
Thu, 10 Jun 1999 15:23:29 +0200


I red somewhere that Linus had problems with installing Linux
on a laptop due to the need of PCMCIA which is not part of
the kernel.

I think what's basicaly flowed about the current way of installing
Linux is that it does not use the Linux special capabilities:
in other words, all current installation programs focus on installing
a single machine from a CD Rom, FTP server or whatever else passive
server.

Nowdays that Linux gets wild spreaded, this is not any more
the only and most powerfull way of installing Linux. Let me ague
a bit about that:
- whatever progesses we may add to the installation program, it will
remain, for years, not usable by the average PC user.
- so providing a more automated program that enables power user
to quickly install a great number of machines or even small companies
to sell a service such as 'bring me your machine and for e few dollars
I will install Linux on it' would help speading Linux much faster.
These companies need a more powerfull installation tool that help
them automate the task with HANDLING the details of each target
computer.

I've build a Pliant (see http://pliant.cams.ehess.fr for details about
Pliant) script that performs the installation the following way:

0) I have a small database of all the machines that I manage, that
contains all their key caracteristics for installing (the processor,
the devices, the size of the disk, the usage of the machine, ...)
which enables the master computer to build automaticaly a customised
kernel for each machine and select the packages or tar files to put
on it.

Now the real installation is performed like this:
1) the master machine build a virtual image of the target machine
on it's own disk (during that step, a kind of preprocessor is applyed
to various configuration files in order to customise them automaticaly
according the definition of the target machine in the database)
2) the master machine builds a floppy containing ONLY the Linux
kernel for the target machine
3) I boot the target machine from the floppy, and it get's the full
environment from the network through NFS root.
4) The target machine runs the script that backups it's disk on the
master, then formats it and install all the required RPMs

How is PCMCIA troubles solved ?

Simply using a DLink DE-620 CT (Ethernet adapter connected to the
parallel prot) in order to achieve the NFS root capability at
install time.

Lastly the installation scripts deals automaticaly with the RPM
limitation: when an RPM needs a given file, it deduces from an
automaticaly generated database, which other RPMs are required.

So, finely, I have the core scipt which is by far the most powerfull tool
I saw for installing Linux (and upgrading also, because it will also
install a new kernel release and genarate all the customised kernels for
all machines through a single command), I am ready to give it to the public
domain in order to help Linux speading, but not to maintain it because
Pliant core development takes all my available time, so if any of you
agrees to be the maintainer of that script, let me know.

Regards,
Hubert Tonneau

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/