Jumping in against my better judgment ... The /64 boundary was for a number of reasons, the fact that "only" autoconfig breaks when that isn't the case is irrelevant (and not entirely true, but many of the breakages are minor/not intractable). Complaining about it now doesn't help, and many other decisions have since been made that rely on /64s. So, half-assed or not - this is the protocol we have, and it works today ... So what is the operational debate? /TJ ------Original Message------ From: Ricky Beam To: Jack Bates Cc: nanog list Subject: Re: Where to buy Internet IP addresses Sent: May 5, 2009 16:53 On Tue, 05 May 2009 16:13:05 -0400, Jack Bates <jbates@brightok.net> wrote:

Actually, they probably would have stuck to a 64 bit address space and it was debated. Then it came down to, let's make it a 64 bit network space, and give another 64 bits for hosts (96 bits probably would have worked, but someone apparently feels the next bump from 64bit is 128bit so there we go).

Ah, but they half-assed the solution. IPv6 makes no distinction between network and host (eg. "classless"), yet SLAAC forces this oddball, classful boundry. Routing doesn't care. Even the hosts don't care. Only the tiny craplet of autoconfig demands the network and host each be 64bits. That's brilliant! Sent from my Verizon Wireless BlackBerry