[arin-ppml] "Millions of Internet AddressesAreLying Idle"(slashdot)

[michael.dillon at bt.com]

> can you explain why someone would stop using IPv4 if it still 
> reached the entire set of endpoints they wanted to exchange 
> traffic with?

That is precisely my point. If someone is using IPv4 addressing
in a private internetwork, or to run their MPLS core network,
they have a restricted set of endpoints with which they want
to exchange traffic. In these sorts of scenarios it is possible
that many organizations will want to avoid migrating away
from IPv4. This is only realistic because these are
LIMITED GROWTH scenarios. In the general case of the public
Internet, growth is continuous with no clear limits in sight.
As the public Internet transitions to IPv6, the IP address
stewards should make sure they keep track of IPv4 blocks that
are freed up, because there could be many, many years left
in IPv4 internetworking.

> right now there's V4-only and dual-stack but there is no 
> V6-only since it offers inadequate connectivity.

That's only true on the public Internet and only if you
discount the bits of IPv6 at the edge which use 
transition/tunneling mechanisms to reach the v4 Internet.

> we can 
> expect a lot more dual-stack as we approach and then pass 
> runout day.  some day most of us expect the cost of V6-only 
> (which will be some moderate loss of connectivity) to be 
> lower than the cost of continuing with (which will be 
> double/triple NAT, high prices for leasing IPv4 RTU [perhaps 
> as bundled with carriage], and router cost and heat for the 
> pea-sized gravel that the IPv4 routing table will become).

I don't believe that dual-stack will be used for very long 
outside of ISP core networks. Businesses are used to technology
refresh cycles, and in a couple of years we will be coming out
of the financial crisis with pent up demand, IPv6 services 
being announce every day, and a pretty clear vision of the 
road ahead. Many enterprises will go to pure IPv6 networks
wherever possible with v4 only running on legacy services 
that are being driven until they drop. Similarly, when
consumers are offered IPv6 broadband it will have v6 gateways
bundled in. In 4 years I am already on my second broadband
gateway device and I would have been on the 4th if I had kept
up with all the upgrade offers.

> but noone i know expects that there will ever be a time when 
> one set of users is releasing IPv4 because they're happy with 
> IPv6-only and IPv4-NAT, while another set of users still 
> needs to grow their IPv4.  unless it's you, that is.  i'd 
> like to hear more about this world view.

I am not sure that there will be a set of IPv4 users that needs
to grow their usage, but I do know that there are an awful lot
more ways of using IPv4 technology than most ISPs are aware of.
Embedded systems, factory automation, spacecraft. Some countries
might profit by sticking with v4 and using our castoff devices,
in the same way some African countries reuse our eyeglasses.
I can't predict this, but I think it is reasonable to be prudent
until we can see that IPv4 really is defunct.

--Michael Dillon