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

[michael.dillon at bt.com]

> what does it take to make your enterprise dual stack?  

Where will your enterprise get the IPv4 addresses required
to go dual stack?

There are many companies whose networks are not growing and
for these companies, dual-stacking is a possibility although
it does tend to lock them into not growing their network. But
for ISPs, whose network is constantly growing, and for whom
growth is a part of the network model, dual-stack is a risky
approach. Of course, if you lose IPv4 customers then you can
use those addresses to continue growing your network but this
cannibalisation strategy is complex and risky.

If a network operator has already moved to MPLS in the core, or
is in a position to implement MPLS without great pain over the
next couple of years, then they can painlessly implement IPv6
services using 6PE with no requirement to dual-stack. And even
if a network does not go to MPLS, you can implement something
similar to 6PE using GRE tunnels or PWE3 to interconnect IPv6
edge routers across an IPv4 core. These strategies using 6PE,
GRE or PWE3 allow you to limit the impact of IPv6 to only those
PoPs and devices that are needed to provide IPv6 service. 
This makes it easier to incrementally add IPv6 to the network
as and when needed.

> we went in early, in fact the earliest commercial roll.  and 
> we have regressed, i.e. some services are not maintained over 
> ipv6 that used to be because no one used them and they caused 
> pain.

We have done similar, i.e. shutting down the UK6x exchange because
there was no business case to keep it going. Also because it was
research and there is no need for basic IPv6 networking research
any more. It's ready for production and we had customers wanting
to buy commercial IPv6 services, so we are in the process of 
migrating the IPv6 services to a proper commercial footing.
Right now the demand is for IPv6 access to the IPv6 Internet
so that is what we are providing. The demand is not there to
provide SMTP over IPv6 so we don't bother with that right now.
I expect that within a couple of years there will be customer
demand for email delivery over IPv6 and we will provide some
kind of SMTP proxy service with DNSBL etc.

A lot of people are complaining that IPv6 is a problem because
there are still some IPv4 services that are not usable over
IPv6 yet. This is a red herring, since these kinds are services
are easy to make available, once the demand is there.

--Michael Dillon