On 2 May 2001, Nathan J. Williams wrote:

> I think that if we're not going to go insane there has to be only one
> way for a driver to attach itself, and that way should work both for
> LKMs and prelinked drivers. We don't want LKM and non-LKM versions of
> drivers to diverge at all.

Oh, there will be just one attach routine. The question is how do we go
about matching when loading a driver that matches existing yet
unconfigured devices.

Take care,

Bill