[arin-ppml] Linking IPv4 allocations to IPv6
jmaimon at chl.com
Wed Jun 25 10:36:05 EDT 2008
Lee Dilkie wrote:
> Seriously folks.
> Dual stack is the only reasonable way to both get IPv6 deployed *and*
> maintain legacy networking with IPv4-only hosts/applications.
> Milton L Mueller wrote:
>> Time to change the header, folks.
>> Just some observations: The idea of linking v4 allocations to v6
>> deployment is a departure (gasp) from "needs-based assessment." You are
>> saying that people can demonstrate need for v4 addresses and not get
>> them if they don't also deploy dual stack. Seems also to require heavier
>> operational monitoring by ARIN. How would one define "gets used?"
> Essentially, yes. If you need IPv4 and don't want to also deploy IPv6
> with it then you are ignoring the common good. Future IPv6-only hosts
> will be unable to communicate with your newly deployed IPv4 hosts.
Aside from the many reasons why this would never fly and likely isnt a
very good idea in the first place, it doesnt accomplish the stated
objective or anything else positive for that matter.
The only way an ipv6 only host can communicate with the entire network
of interest is with translations and gateways. This is because it simply
hasnt and wont happen anytime soon that every ipv4 node of interest will
also have an ipv6 address.
I would lay odds it isnt going to happen for 5-10 years either, if ever.
No matter what runs out.
Do you even think there are enough netop man hours to ipv6 every single
network and node starting from today until IANA runout?
So dual stack is a dead end concept that hasnt and will never fly. It is
only required today because ipv6 doesnt actually connect you to a
network with anyone of interest on it.
Dual stack is a policy of failure. It is a failure to make ipv6
compelling and useful on its own.
Not only does ipv6 need to be compelling and useful in its own right, it
must also provide at least the same level of network access as v4nat.
And for a nitpick, any such policy would actually have to require that
all addressed nodes from the allocations be dual stack to have any hope
for a change at all. So the policy would have to require that the
requesting org dual stack all hosts to be addressed from the new
allocation and get ipv6 transit, which last I checked is barely
available most places.
As for policy driving IPv6, I think that ipv6 allocations should be made
available for the taking for every single asn originating ipv4 prefixes
and it should be large enough that they will likely not come back again.
A /32 for every AS. No questions asked. We have at least half a billion
to hand out, I am sure we could spare 40-50k. In fact, the allocations
should come from a range that includes the AS number.
Once they have it, they might use it. Tunneled or not.
But making ARIN into any more "bad cop" than absolutely necessary is a
Really Bad Idea.
More information about the ARIN-PPML