[arin-ppml] Policy Proposal 2008-6: Emergency TransferPolicyforIPv4 Addresses - Last Call

[michael.dillon at bt.com]

> Here's the interesting dynamic; there is almost no 
> first-mover advantage to deploying IPv6 first.

This is not entirely true. There are now large organizations
that have decided to deploy IPv6 for strategic reasons and
that want one or more network operators to supply them with
IPv6 services in order to assist them with their tests and
trials. The number of such organizations is on the increase
and this presents the "first mover" ISPs with the opportunity
to break existing supplier relationships with their competitors
who do not offer IPv6 yet.

> However, beyond that it doesn't get you 
> extra revenue, and may make you purchase equipment you would 
> not otherwise purchase, run newer software with more bugs, etc.

Existing network hardware (servers and routers) has had IPv6
support for years now. Nobody is seriously suggesting that there
is a need to buy new equipment beyond the normal technology
refresh cycle. As for software, I don't believe that there is
a correlation with newness and bug counts. In general, software
packages continually get newer features added whether you use
the features or not. This is what drives bug counts, and you
will not escape this by not deploying IPv6.

> We see ISP's state this over and over, the customers do not 
> want it, so we're not doing it.

But when the customers do start asking for it, will these ISPs
be ready with a tested service, or will they start losing those
customers to their competition?

> ISP's will also move to push the transition costs 
> to their competitors, last one with a transition box looses.

Makes more sense to push that cost onto customers because
you have less issues of scaling if the customer has the
transition box on their premises.

> Akamai's and 
> Limelight's I suspect could offer IPv6 services worldwide in 
> a matter of a few weeks, if there was a reason to do so.

Yes, and there are a couple of companies working on virtualized
IPv4 machines with IPv6 translation bundled in. Basically, they
copy your server's hard drives into a virtual machine hard drive
running XEN or VMWare, and then the host machine layer transparently
does the NATPT or NAT64 services transparently to the legacy
v4 server which has now been virtualised.

> In short, the least effort, least cost path is to delay as 
> long as possible, then for everyone to leap forward at a 
> brisk pace; essentially as fast as possible without paying 
> money just to "speed it up."

No argument there, just that everybody has to be deploying IPv6
today in their test and trial facilities to make this work. Most
of the big companies are doing so, but we need to get more of 
the mid-sized ISPs doing this as well.

--Michael Dillon