> What's being suggested here is taking code that belongs in userspace
> and putting it into the kernel.  Code to govel namelists and dumps of
> different kernels simply doesn't belong in the kernel.

And this is basically where the opinions differ - some of us think the 
paradirgm of "everything is a file" is a lot nicer than the kvm system. There 
is of course nothing preventing one from implementing both. Linux e.g. has 
both a /proc which is basically a nice interface to the kernel functions.

Peter