MAC vs DUID issue

[sthaug at nethelp.no]

> There have been "vigorous discussions" before about MAC vs GUID ! My understanding was that the MAC is not delivered in the DHCPv6 packets and are not available to the server to use. The reason for this is that those who designed the spec felt that MAC addresses have several flaws - there can be more than one per device, and they can change due to hardware changes. The counter point (to which I subscribe) is that while MACs have issues, they are for many the least bad option since they are fairly invariant, remain across multiple OSs and OS reinstalls (and in your case, 'careless' cloning).

I agree with this point of view too - it was a serious mistake to drop
the MAC address.

> IIRC there was talk of defining an option to carry the MAC address, but I can't remember what the outcome was, and in any case it could be "some time" (if ever) before all devices send it.

RFC 6939 (DHCPv6 Client Link-Layer Address Option). Unfortunately, I
don't know of any equipment which implements this yet. Also, note that
this option is *only* meant for relay agents. On other hand - if your
DHCPv6 client is on the local LAN, you can of course look at the source
MAC address directly...

There's been some spirited discussion about this on Nanog recently,
see for instance

http://mailman.nanog.org/pipermail/nanog/2014-March/065726.html

Steinar Haug, Nethelp consulting, sthaug at nethelp.no