In some email I received from Christoph Badura, sie wrote:
> darrenr@reed.wattle.id.au (Darren Reed) writes:
> >If NetBSD were to support long disk device names which were actually
> >addresses to the device, it'd just mean it were possible to be even
> >lazier in using NetBSD.  It would also require some extra magic - be
> >it a special filesystem type or running a configure program at bootup.
> 
> I find it kinda amusing that noone has mentioned the "put in another
> controller and suddenly /dev/dsk/c0tNdMsO becomes /dev/dsk/c1tNdMsO"
> problem yet.  ISTR that the last time this carcass was beaten to pulp
> that was the the argument that shot down the whole idea because you
> have to wire down the controllers to the busses, which was decided to be
> impractical for "generic" kernels.

And that depends on which `slot' your controller gets put in, but yes,
point taken.  But what happens more often - changing controllers or
changing devices ?

Darren