[ppml] Policy Proposal 2005-1: Provider-independent IPv6 Assignments for End Sites - Last Call

[Vince Fuller]

IMHO, this policy is a mistake.

Quoting from the policy proposal:

> This policy proposes:
> 
> o Large and/or multihomed End Sites receive assignments directly from ARIN.
> 
> This policy applies to organizations with networks that are large and/or 
> multihomed. Like their IPv4 counterparts they do not make assignments to 
> external organizations. They instead assign space internally to their 
> own facilities. Similarly to IPv4 These internal assignments are not 
> submitted to ARIN via swip/rwhois.

It is precisely the "large and/or multihomed End Sites" that are driving
growth of IPv4 routing state to be super-linear. Adopting this policy is
a giant step toward non-scalable routing by the creation of an ipv6 routing
"swamp" that will make the IPv4 global routing situation seem positively
arid by comparison.

Implementing this policy in the hopes that it will spur adoption of ipv6
smacks of the "tail wagging the dog", of doing *something* rather than the
*right* thing so that "progress" can be demonstrated. Such an illusion of
progress will only alleviate pressure to fix the real flaws in ipv6 (i.e.
co-mingling of the endpoint identifier and routing locator in a single
"address" field) that render it incompatible with the goal of a scalable
routing system.

Keep in mind that exponential and quadratic growth curves ramp-up very
slowly when initially dealing with small numbers. I fear that by the time
the consequences of this policy decision are felt, by the time the ipv6
routing state growth curves are observed to be super-linear, it will be
difficult or impossible to deploy a something to handle that growth. Few
people realize or remember how close the Internet routing system came to
collapse during the pre-CIDR days...

	--Vince