For a feel for some of reasons your question sounds naïve, see Geoff
Huston's presentation on IPv6 and CGNs at APNIC 32 all the way back in 2011
(trust me, it's entertaining, as Geoff usually is):
Slides: <http://labs.apnic.net/presentations/store/2011-08-30-exhaustion.pdf
Recording: <http://webcast.apnic.net/meetings/32/opening-hinted.mov>
On Fri, Nov 22, 2013 at 11:59 AM, Ted Lemon
<ted(_dot_)lemon(_at_)nominum(_dot_)com> wrote:
On Nov 22, 2013, at 11:46 AM, Noel Chiappa
<jnc(_at_)mercury(_dot_)lcs(_dot_)mit(_dot_)edu>
wrote:
CGNs are expensive. Why would people prefer to maintain them if the
IPv6 infrastructure was working?
Cheaper than customer service calls.
That's almost certainly true now. I'm not talking about now. Really,
the big problem with CGNs is that they don't scale—this is why lw4over6 and
MAP represent a significant and useful innovation.
What I expect to see happen in the market is that in the next decade, CGN
will start to seem old-fashioned, and more and more what will be deployed
will be stateless port-sharing NAT hardware in the core, supported by
stateful NATs at the CP, just like we were doing before the CGN idea got
popular.
The nice thing about stateless NATs in the core is that they can be scaled
up and down according to need. So as more traffic is carried over the
native IPv6 network, you can start switching off boxes in the core. You
can track the demand, so you don't switch off boxes that are needed. You
can even goose the switchover by running your IPv4 NATs closer to capacity,
forcing traffic that could go over either transport to go over v6, because
it's working better, and Happy Eyeballs can tell.