[ppml] [address-policy-wg] Those pesky ULAs again

[Paul Vixie]

> > ... and i want to be able to use non-sparse addressing, such as a /112
> > (65536 possible hosts) rather than a /64 (which is 18446744073709551615
> > possible hosts) per subnet.
> 
> But it seems usable or at least could be improved/built on.  Is there a
> reason not to have at least /80 on a 48 bit MAC layer?

leo touched on a number of good reasons for not wanting to always base one's
network address upon one's transport address.  the only reason ietf reached
consensus on the EUI strategy was that at the time, these addresses were
expected to be short lived, renumberable while in motion, multihomed most of
the time, and so on.  once they lost their ephemeral nature, we should have
done a nancy reagan ("just say no") to the EUI approach.  (so, more guilt.)

> For EID/RID, well who knows what form it will eventually take.

in all likelihood, the network protocol for split EID/RID won't be IP at all,
or at least, not IPv4 or IPv6.

meanwhile, to urge things back toward the topic at hand, we need policies for
effectively and efficiently allocating IP numbers, including IPv6 numbers, and
i remain convinced that asking any RIR to measure the "routability" of these
numbers is a recipe for disaster to be prepared while running on a slippery
slope carrying scissors.