Re: kernel 2.4.x and unstable/ Kernels
Nathan E Norman wrote:
> On Sat, Sep 08, 2001 at 09:42:34PM -0700, Craig Dickson wrote:
> > Nathan E Norman wrote:
> >
> > > Once you manage more than one machine, you will find reasons to prefer
> > > kernel-package :)
> >
> > You mean, once I manage more than one machine that will run exactly the
> > same kernel. Right now, I have two Debian boxes, but one is a P3 running
> > Sid (kernel 2.4.9) and the other is an original Pentium (complete with
> > the FP divide bug) running Potato (2.2.19).
>
> Uh, this doesn't make sense. There's no limitation which says you
> can't build kernels on one machine which are customised for another
> (entirely different) machine.
True, but as my first sentence above makes clear, I was assuming that
you meant to build one kernel and install it on a number of machines.
The rest of your latest message clarifies that you actually meant that
you want to build all your kernels on one system and then distribute the
resulting packages out to the other machines. To which I say, sure, but
kernel-package still doesn't seem all that advantageous. All you have to
do is copy two files (vmlinuz, System.map) across the network and
replace the /vmlinuz symlink, so how much work do you really save by
copying and installing a kernel package instead?
> The p90 and especially the 486 are
> practically useless while compiling a kernel.
I buy that for the 486, but my P90 builds its own kernels. It's
definitely a lot slower than the P3-700, but since the P90 exists, at
this point, solely for DMZ proxy duties, I don't really care how long it
takes.
> Since the athlon has
> the biggest (and fastest) disks, I need kernel-source trees on one
> machine rather than spread all over the place (let's see, I downloaded
> the ipsec patch where? Did I patch this tree, or was that the other
> box?) This is a significant advantage.
Yes, but it's not relevant to whether you use kernel-package or not, is
it?
Craig
Reply to: