[arin-ppml] Preemptive IPv6 assignment
On Wed, Oct 13, 2010 at 9:39 AM, George, Wes E IV [NTK]
<Wesley.E.George at sprint.com> wrote:
>>I claim that anyone actually announcing routes with a registry AS
>>number has demonstrated a defacto need for registry IPv6 addresses.
> [WES] No, having a registry AS# and announcing
> routes != having an allocation from $registry. It simply
> means that they are multihomed. There are plenty of networks
> who get PD IPv4 from one or more upstreams and announce
> it to all of them via BGP.
> I acknowledge that multihoming with PD space in
> IPv6 is not necessarily attractive due to risks of your
> subnet announcements being filtered on some networks,
It's not really a question of attractiveness Wes, it's a question of:
"does it work?" From what Jason Schiller was telling me, it appears
that about 50% of IPv6 ISPs filter /48 customer cutouts from
ISP-allocated space. If his estimate is correct then the answer is:
No, multihoming using one of your ISPs' space _does not work_ in IPv6.
What does work, now that Verizon is on board and accepting /48
announcements, is a registry IPv6 assignment. So if you're multihomed
and announcing IPv4 prefixes, it's little short of a certainty than
when you deploy IPv6 you will be multihomed and announcing and
announcing a registry-assigned IPv6 prefix.
> but it's not impossible, nor is that germane to the discussion.
Respectfully, "does it actually work" is germane to -every- policy discussion.
> I simply don't see the point at which this helps spur deployment.
Does the quarterback refuse to throw the ball because no receiver is
in a position to make a touchdown? Of course not. Let's get the
allocations and assignments we know are needed in to the network
engineers' hands. Unless we're ready to punt to the v4 CGN team, we
need to move the v6 ball down field.
William D. Herrin ................ herrin at dirtside.com bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004