Re: Path to kmods

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Thu, 13 Nov 2008 18:49:38 +0100
    From:        Johnny Billquist <bqt%softjar.se@localhost>
    Message-ID:  <491C68B2.5090601%softjar.se@localhost>

  | Don't we get into a chicken and egg situation here?
  | The FS is a module, but we need the FS in order to load the module.

Maybe, maybe not, but that wasn't the point at all - for my example
you can assume that the particular system only uses NFS once a week
for mounting a pkgsrc binary packages repository if you like.

The actual issue related to whether it is going to be possible to have a
single (common) module that works on a particular architecture/version kernel,
regardless of the options with which the kernel was compiled - for all
reasonable potential modules, and for all kernel compilation options.

If the answer turns out to be "no" then modules really need to be attached
to a particular kernel.

In a way, the question is the same as why, when compiling kernels now
(all the various kernels for a particular version, and architecture)
we recompile everything from source rather than sharing many (or most) .o 
files (perhaps from a .a file).

kre