renumbering

[Jun-ichiro itojun Hagino <itojun@iijlab.net>]

>> If NAT works so well in the first place, why do we then need IPv6 at all?
>
>i expect that question to come up quite a lot, as part of the business case
>analysis.  "i've got 20,000 ip-speaking devices, but only 7 of them have
>universally unique addresses, the rest are in private address space because
>i was not able to get a class B.  if i move to ipv6 i can get my own /48 but
>it will be in my isp's address space, which will make multihoming hard and
>make switching providers even harder.  do i really want to have to renumber,
>even with ipv6's extensive automated assistance for this, whenever there's
>a rate war at my local transit exchange?"  and then will come the question
>you gave above.
>
>(i'm done grousing about A6 -- that's a dead issue as far as i'm concerned.
>but the problems A6 was supposed to solve or avoid didn't die with it, and
>we're going to have to grapple with the business case analysis every step
>of the way to full ipv6 deployment.  unfortunately, the early adopters are
>all organizations of the kind who get their address space from RIR's rather
>than from their upstream ISP's, so this issue hasn't yet seemed real here.)

	i have renumbered my home (/48) more than a couple of times without
	any problem.  yes, it may not extend naturally to enterprise network 
	where # of nodes are 10^4, but anyway,
	- IPv6 address autoconfiguration
	- multiple IPv6 address per interface
	- source/destination address selection
	makes renumbering so much easier than IPv4 case.  you just need to be
	careful about DNS TTL and RA address/prefix lifetime.

itojun