[arin-ppml] Policy Proposal: Dedicated IPv4 block to facilitateIPv6 deployment

[Paul Vixie]

> > what do others think?
> 
> My personal opinion is that we have to make policies using the best
> *current* knowledge of how routing works. Filters on prefix length exists
> *today* and we must take them into account. If those filters go away in the
> future and the internet goes to flat route every single /28 or so, that is
> about 250 million routes... We all would be in a much different place.

under that reasoning, would there also be no IPv6 PI?  would there be any PI,
even for multihomers?  would address blocks only be handed out to folks with
intercontinental backbones whose intent was to use this address space for
their customers?

if not, why not?  where are you choosing to draw the line and say, here we'll
care about how routing works, but beyond this we won't?  current policies and
current proposed policies including relaxed transfers will yield a megaroute
core years before that's cheap enough for current tier-1 and tier-2 let alone
tier-3 and enterprise multihomers.  and i don't actually feel all that sure
that a megaroute core would work even if everybody could afford the CAM; that
size DAG with no hold-down and no hold-up and current churn rates scares me
and will scare me until some math weenie graph theorist explains why it would
be ok.

> Knowing what we know about the current generation routers, it seems prudent
> to craft policies that not only do not have too negative an effect on
> routing table size, but also care about ACL size.

do you want folks to be able to say, "hey, you should accept our route, it's
in an arin policy?"  and do you think this will make some difference?  and yet
you think the industry should resist the siren call of revenue that comes from
/28 multihomers hoping to avoid provider lock-in in a post-iana-pool world?