After 13 years of slow, the fact that a tsunami of v6 is about to overrun us may seem surprising. However, the next 2 billion internet users and the next billion smartphones aren't going to be on v4, because native v6 will be cheaper for them than carrier NAT444. Anyone dealing with asian customers, asian supply chains, mobile devices, or government contracts is going to need v6 Real Soon Now.

ISP's who don't offer any public IP's (private double NAT444, no v6) are going to lose business to ISP's who dual stack, because the NAT444 costs will be higher than v6 costs, and customers will hate not being able to do streaming video or gaming over NAT444, and businesses will hate not being able to geolocate customers.

It's like the transition from analog TV to digital TV. Much of the same old content, but new gear need to transmit and receive it, and taking multiple years to deploy. The consumer tipping point may be sooner than you think - I propose a v6-only asian electronics toy around Christmas 2014, with 99% v6 traffic around 2017, and a tier-1 backbone IPv6 flag day where IPv4 routing is dropped in 2020. The post-tipping point economic incentives for ditching v4 routing are very strong, unlike the disincentives for starting the initial v6 rollout.

A killer problem with NAT-PT, at least if you are talking ISP carrier grade NAT, is that the NAT gateway has to fake an client v4-only A DNS record for the v6-only AAAA server. You can't do this reliably at internet scale. You can't use very much v4 space for the mappings, or you lose connectivity to chunks of the v4 internet. But you have to use lots, because thousands of customers are opening hundreds of simultaneous connections to v6 web sites. Now try to get the lifetime of the mapping right in the face of client-side caching, persistent HTTP connections, load balancers, round robin, and content distribution networks. If you guess too short the persistent connections die. If you guess too long the next customer connects to the wrong place. Oh, and your fancy NAT46 device has to be less expensive than 6rd (a v6 over v4 tunnel from your broadband modem to a relay at the ISP).

Good luck with that, you are going to need it.

If you ask John Curran, CEO of ARIN, and current cheerleader for IPv6 due to lack of any good alternative, he *agreed* with DJB about the economic disincentives and lack of transition planning. This is precisely *why* IPv6 just spent 13 years (1998-2010) wandering in the wilderness, unadopted. However, the "mistake" was made in 1993, before the major commercial growth of the v4 internet, when the main transition to analogize by was the 1981-1983 ARPANET cutover from NCP to IPv4. Which took multiple years, and had a messy interval where the quality of the v4 network stacks was mediocre. It should be unsurprising to see those characteristics replicated in the v4 to v6 transition.

Note that today you can reach google and cnn over v6, though probably not yet your own firm, and still less than 20% of the Alexa top 100 or 1% of the top million web sites. Consumers in the US should continue to ignore v6 for 12-18 months while ISP support, broadband modem support, and wifi router support improves. But ISP's, businesses with asian supply chains or asian customers or mobile customers or government contracts had better get busy rolling out v6 on their outward facing services.

Actually, the forthcoming "4G" smartphone rollouts in the US *are* IPv6 native, with carrier grade NAT44 for the legacy internet, exactly as you advocate. This is already happening.