[ppml] "Who's afraid of IPv4 address depletion? Apparently no one."

John Curran wrote:

> Without IPv6-only users or some compelling benefits for IPv6 versus IPv4
> transport, it's hard to see why the content community would invest ahead
> of time.  Existing enterprises, consumers, and content providers all have
> working infrastructure that still meets their needs once there is no readily
> available free pool; it's the ISP business growth model that gets impacted
> and hence the ISP community that needs to drive any desired transition.

In the R&E community, there is still a credible belief that we will see 
IPv6-only resources in the fairly near future.  For example, the Large 
Tokamak project in Asia is reputed to be IPv6-only once it becomes 
operational.  Researchers elsewhere who need data from it will need v6 
or some sort of proxy.

I do believe that we will see IPv6-only users in parts of the developing 
world, where they won't be able to afford increasing prices for IPv4 
addresses once the free pool run-out occurs and we (maybe) move to a 
market system.  In the R&E community, there are a bunch of us who 
believe we need to make our resources available to that segment, but, 
like any other sector, do not speak with one voice.

I am glad to see that John has had good luck in his conversion, but I 
agree with Randy that there are many bumps in the road.  See:

http://events.internet2.edu/2008/jt-hawaii/sessionDetails.cfm?session=3607&event=278

(video stream at http://winmedia.internet2.edu/jointtechs-w08/jtw08-18.wmv )

My point in that talk is that there are bumps even in places where you 
don't expect them, such as applications that _do_ support IPv6 (but not 
easy dual-stack) or registrars that, to their credit, have implemented 
AAAA glue for years (but had some bugs in their implementation and there 
hasn't been sufficient use of the registration system to tickle those 
bugs).  However, for me and my organization, that has been justification 
for moving ahead now.  It's precisely because with think there WILL be 
some pain, bumps, and delays in converting to dual-stack that we feel we 
need to get started now.  We still have a little bit of time before the 
crunch comes; if we move now, we believe we will be in a much better 
position than if we wait until we absolutely have to migrate.

As for transitions, given the cloud that now hangs over NAT-PT (via RFC 
4966), and Iljitsch van Beijnum's point in "Running IPv6"--that 
applications that support IPv6 will tend to have ALGs and proxies that 
those that don't have ALGs and proxies won't support IPv6 from the 
client side anyway--I am moving toward a combined v4-NAT and v6 model. 
My idea (for large workstation lans and residence services) is to have a 
one-armed NAT gateway to do the v4 translation, while v6 would be just 
routed through the regular routing infrastructure, with no middleboxes. 
  I am in the process of a POC now.  (If anyone else has tried this let 
me know off-list, as that's more of an ops issue.)  This still requires 
clients to be dual-stack and it still has the yuckiness of v4 NAT, but 
it does give clients an end-to-end transparency option (and thus an 
incentive to use IPv6) plus a v4 fallback option, and it could reduce 
pressure on IPv4 address space in the future.  It seems like it might be 
useful on large client networks, WiFi nets, residential nets, etc.

michael