[ppml] Proposed Policy: IPv4 Countdown

[Tony Hain]

Edward Lewis wrote:
> At 8:37 +0000 3/15/07, <michael.dillon at bt.com> wrote:
> 
> >In my version of this thought experiment, the ARIN region quickly
> >realized that roughly 90% of their infrastructure was capable of
> running
> >IPv6 with only a software upgrade so they did upgrade. Problem solved.
> 
> What I am hearing and experiencing is that this isn't happening.

The crisis has not occurred yet, so why would you expect any significant
level of activity?

> 
> No one has a goal of using IPv6.  I've never gotten a budget request
> approved "just to update the technology."  No matter how trivial or
> cheap, change costs something, and more significantly, represents
> risk.  A lot of the lessons I learned involved the introduction of
> 10BaseT on a government campus, i.e., it's been a while.  Have
> accounting people become that much more friendly to IT?

No, they have not. This is clearly a cost/benefit trade-off to the
accountant. When the price on the contract from the ISP for retaining IPv4
exceeds the cost of training/testing/etc. for the IPv6 deployment, the CIO
will follow the path of lowest cost to keep the accounting people at bay.
Until then there will be resistance because as you note, change has a cost.
For large organizations with PI space that date is likely to be far into the
future, assuming they have sufficient space to accommodate growth (else they
end up in the open market). The driver for them is more likely to be that
their suppliers are smaller and have to live in PA space, so these suppliers
will be clamoring to shift to IPv6 by claiming they will have to raise their
prices to the business partner to cover their rising costs. 

The feedback mechanism that will drive business decisions is cost/benefit,
so until the cost rises there would have to be some application of clear
benefit. Those applications can't make it to market in a world that is
dominated by nat, because the dev/support costs are just too high. That is
why 7 years ago I put a stake in the ground and said that Windows would
tunnel and create a virtual IPv6 network so that the app development
community would have an environment to deploy into that was free of concerns
about nat awareness & traversal. That tunnel model will persist until the
service providers offer a native IPv6 path, at which point the stack will
prefer the more efficient route. 

Tony