NPT66 is NOT a solution
So, more FUD about IPv6 with an article that totally fails to grasp key IPv6 concepts.
1) NPT66 is still hardly more than an RFC, there are *no* trivially accessible implementations of this (as in, for the low-end folks), and sure as shizz not in home routers or even the latest build of OpenWRT - the only way you're going to get it is by merging a bunch of currently unstable Netfilter patches into the Linux kernel source and building yourself.
2) It's largely unnecessary anyway - ISPs can delegate prefixes of /64 or larger to clients (either through RAs and/or DHCPv6-PD) which could then be announced on the LAN side for assignment to clients (something that could be achieved with ISC'd DHCP client and/or a bash script to invoke radvd and optionally dhcpd)
3) Don't confuse multihoming with poor-man's load balancing involving round-robin SNAT on multiple separate IPv4 addresses - exactly the same tosh can be done with IPv6 but the responsibility moves to the endpoint (i.e. you give a machine an IP in every one of your subnets and configure it to use them in some per-connection rotated fashion) - of course, I have no doubt the plebiscites will be utilising round-robin IPv6 SNAT once it gets mainlined into the kernel.
4) Suggesting the use of BGP to be a bad idea because of an issue in China is mentally retarded when you take a moment to that your provider, or their provider MUST BE USING BGP since, y'know, it's the backbone protocol of the *entire* Internet and therefore, any upstream prefix hijacking is basically *unavoidable* - on the contrary, at least if you do BGP yourself, you have the option of using stuff like pgBGP to at least have a chance of handling prefix hijacks.
5) You can actually get a free IPv6 BGP tunnel from companies like HE providing you have your own ASn and subnet assignment from an RIR which is generally affordable if you get it via a sponsoring LIR, but also only something either an enthusiast or small business would do.
Biting the hand that feeds IT
