[ppml] IP-v6 Needs (RE: a modified proposal 2005-8)

[Michael.Dillon at btradianz.com]

> I'm not disagreeing, but what is the definition of "end site" here?
> The aircraft has a single /48? 

Yes, the aircraft is the subscriber. Of course many aircraft
will be owned and operated by a single company, but not all
of them.

> Are you also saying that each /48 be provider independent? 

No. The design of IPv6 says that the end-site uses /64 subnets
within it's own network topology. Those /64 subnets allow each
subnet to use provider independent addressing for the host-portion
of the address, for instance they could use Ethernet MAC addresses
as the stable portion of the address. For this to work, they need
a fixed and predictable number of bits for working space, first
for the host addressing and then for the subnetting in their
own topology design. This subnetting is also intended to have
space to evolve so that most organizations only need one allocation
from either and RIR or a provider, during their lifetime.

> Buses, trains or train cars, ships, tandem bicycles.  Be
> specific.

I think that being specific is contrary to the goal of 
coming up with a general policy that is not unnecessarily
restrictive. It doesn't matter how the networks move around
and what physical objects they are attached to. If the discussion
gets to specific it always leads to corner cases which don't
work. I don't believe it is possible to have a policy without
corner cases where the policy doesn't work. In the past our
policies have had issues, and over time, people either live
with these issues or they come up with workable new policy
proposals that improve the policy. This is a good thing.

--Michael Dillon