[arin-ppml] Policy Proposal: Dedicated IPv4 block to facilitate IPv6 deployment

[Paul Vixie]

> ...
> but what is this gonna tell you?  what is in place today.  bit whoopie doo.
> 
> just as with the 2050 /19 agreement in danvers, filtering policies will
> adjust to allocation realities if honestly negotiated and technically
> feasible.
>
> so do not base all future technology and policy on yesterday's ephemeral
> router configs.  otoh, do make tools to prevent, diagnose, and fix errant
> configs.

overall, this is my position also.  we already need some kind of routing
security to prevent injection of prefixes by unauthorized parties, and it
has to be based on public-key crypto rather than prefix lengths or prefix
quotas.  prefix length filtering "on RIR boundaries" is too coarse, and
prefix limits per peer are too coarse.  therefore without any new policy
and without allocating any /28's from anywhere, we already have the need
for technology which, once present, will mean that it doesn't matter where
the /28 microallocations for v6/v4 gateways come from, even assuming that
demand for these is from multihomers who cannot just use PA, which hasn't
been successfully argued (to my satisfaction, anyway).

my concern is, we're not there today.  a policy that said folks could get
/28's for use in v4/v6 gateways might not be attractive or even useful, if
these prefixes wouldn't get good reachability.  and while i agree that a
well reasoned policy in this area would have the effect of changing the
internet's route filtering culture, i think early adopters might feel some
fear, and this fear might be well founded if it takes a while for prefix
length filterers to adjust to the new policy.  so, i'm suggesting that the
reachability of these prefixes be measured and reported, and compared to
the reachability of older, larger, more stable prefixes.  this could be
done not just once, but periodically, and so detect any trends.