Re: Bug#219582: ITP: linux -- Linux 2.4 kernel
This one time, at band camp, Marcelo E. Magallon wrote:
> > How do the current kernel packages guarantee this?
> > Why would Robert's package need to behave any differently?
> The current kernel packages don't make the old stuff just dissappear,
> so it's less of an issue in that case. In fact, the only "bad"
> situation with the current kernel packages is when you update between
> package releases of the same kernel version, and the current kernel
> packages make plenty of noise in that situation.
I've had another thought, which was spurred by the System.map discussion;
and some people are probably going to hate it because it duplicates some of
the effort of having a package management system in the first place.
The grub package doesn't ever install itself to /boot, it requires the admin
to copy the binaries from /usr/lib after an upgrade (or my memory is totally
flawed). It wouldn't be so difficult to rotate the previous (good) kernel
and associated files and replace them with the new kernel.
An update-kernel script which ran after installation, and again at boot
time, could check to make sure the latest kernel was in place and that ones
bootloader could find it, and that the previous kernel was also accessible
to the bootloader.