Re: An initrd proposal
On Mon, Oct 25, 1999 at 12:48:09PM -0700, Joey Hess wrote:
> Florian Lohoff wrote:
> > You have the problem here that you afterwards need a different kernel
> > for booting the resulting installation this might lead to ...
> >
> > "After the installation the system wont boot" ...
>
> Well one workaround would be for debian to use an initrd to boot, after
> installation, too. Of course, experienced users would build their own kernel
> and bypass it.
>
> Is that a crazy idea?
Nope - Redhat does it like that AFAIK - They link the initrd to the
kernel image and let lilo load it with that on startup ... That should
be doable on bootup ...
> I'm very impressed by Nils's message. I've been thinking along similar lines
> for months, mulling over the same ideas though I didn't come up with
> anything as concrete as Nils has.
I had similar thoughts a year ago when i built some boot/autoinstall disks
for a 500 Node Cluster experiment here in Germany.
I modularized EVERYTHING but did only include a couple of modules
in the Disk itself (The ones needed) with some lines of config
which modules to load with which parameters ...
Thinking a bit further -
Have the Kernel always include ALL modules needed for a possible
destination (This might IDE + SCSI) and modularize ALL sources
(CDRom + Network). Then build 2 Disks - One with the CDROM and
one with the Network drivers.
Once you are ready to access the source (Which is the first
you try to do) you fetch all needed additional programs from
the source.
Why not letting choose a debian.mirror from the network
installation disk and let it fetch the base2.1.tgz via http ?
Flo
--
Florian Lohoff flo@rfc822.org +49-5241-470566
... The failure can be random; however, when it does occur, it is
catastrophic and is repeatable ... Cisco Field Notice
Reply to: