host-identifier with IPv6

Peter Grandi pg_dhcp at
Tue Mar 3 09:25:44 UTC 2009

>>> On Mon, 2 Mar 2009 17:04:49 -0700, Ted Lemon
>>> <Ted.Lemon at> said:

> [ ... ] On the other hand, when I get a new Macintosh, I
> typically do a brain transplant from the old Macintosh.  So
> for me, the MAC address changes, and the DUID is stable. [
> ... ]

As an aside, which is related to DHCP in some way, I practice
something similar; I give proper names (that is, DUIDs) to
filesystems as well, because in practice a "system" is really
defined by its persistent state; copy a set of filesystems from A
to B, that's the real "brain transplant" and that means that B in
effect has become A.

After many steadfast years of meditation and reflection :-) I use
in my configuration scripts three env vars called SITE (for the
environment), HULL (for the hardware) and NODE (for the "real",
filesystem-based, identity of the system). A "brain transplant"
then changes the HULL, and attaching to a different network changes
the SITE. Note: this classification could be finer (e.g. split SITE
into ORG and NET), but I found it captures most of what I want.

How does this relate to DHCP? Well, the traditional view of IP
networking is that addresses identify interfaces and not systems,
and I feel that this makes discussions about the merits of
responding based on DUID vs. hw address a bit difficult.

Unless one takes the point of view, which seems popular, that DHCP
really is a general purpose query-response layer2 protocol, and the
semantics of its queries and responses are essentially arbitrary
(and not even necessarily related to net or system configuration).

More information about the dhcp-users mailing list