IPv6 PI allocation

Benny Amorsen benny+usenet at amorsen.dk
Fri May 18 14:28:25 CEST 2007


>>>>> "CM" == Colm MacCarthaigh <colm at stdlib.net> writes:

CM> On Fri, May 18, 2007 at 10:13:38AM +0200, Benny Amorsen wrote:
>> IMHO the assumption of a large amount of hosts sharing a medium is
>> IPv6's largest failing. It will never happen again -- but of course
>> we can emulate it.

CM> Huh? I don't see that assumption anywhere.

Sorry, I should have said "more than one host", instead of "a large
amount". 

CM> The /64 for hosts isn't so much "we expect trillions of hosts per
CM> subnet" so much as it is "we'd like to make host-address
CM> uniqueness trivial, oh and we'd like to be able renumber lots for
CM> privacy reasons without collissions, oh and we'd like to avoid
CM> ever having to renumber a lan just because it was too small at
CM> time of inception".

It's the assumption that there is a LAN at all, that I'm challenging.
LAN's don't physically exist anymore, we just emulate them with
switches. When everything is point-to-point, we're carrying 64 bits of
baggate for purely historical reasons.


/Benny




More information about the ipv6-ops mailing list