>Dynamic DNS just papers over the problem.  Why should the DNS even have
>occasion to care when I change network cards?  In IPv4, this is what
>ARP was all about (at least for Ethernets; analogous remarks apply for
>other media).

	talking about ARP...
	In IPv4 it was required to implement link-layer address lookup for
	different medium.  IPv6 tried to solve it by implementing link-layer
	address lookup into ICMPv6.

>> I believe this problem is is not just for IPv6, it just made it
>> visible.
>Well, if providing less of a solution for it that IPv4 did counts as
>"just ma[king] it visible", perhaps.  With IPv4, I can shut down a
>machine on my house LAN and bring it up again with a different MAC
>address (change network cards, or move the disk to another machine, or
>whatever), and the only difference will be that the ARP caches will
>look slightly different.  I don't have to bash the DNS, I don't have to
>frob the router, I don't have to retrain my fingers (which know the
>addresses of my house machines fairly well), nobody off-segment can
>even *tell*, never mind *care*...it all Just Works.  It sounds as
>though this is not true under IPv6, in which case I can't see it as a
>step any direction but backwards.

	it is not required to use MAC address as source of link-layer address.
	you can swap them with some other number as you like.

	if you prefer to configure it manually, that's okay.  please replace
	link-layer address and you are covered. (we need to check if the
	code is working or not)

	if you want it to be automatically configured, we need some number:
	- that is unlikely to conflict with other node
	- and do not change over time/reboot/hardware swap
	there's not that many values with this property.
	- MAC address? - if you change ethernet card you lose.
	- Pentium333 CPU ID? - if you change CPU you lose.
	- MD5(hostname)? - chance of conflict?  many novice guys leave it
	  unconfigured.
	- MD5(ROM BIOS)? - chance of conflict is very high.
	the default behavior described in the document is to use MAC address.
	I think it was not that bad choice.

	BTW privacy concern adds more interesting issues...
	draft-ietf-ipngwg-addrconf-privacy-01.txt

itojun