The think I find mindboggling about all this is that I have yet to see a
concise explanation of how the great transition to IPv6 is going to be
managed and what the incentive for early adopters is going to be.
Nostalgia. :-)
If you find yourself yearning for a simpler time and place, where spamming is unknown, nameservice is still a novelty and people are more interested in the network itself than in the content it may carry, IPv6 is the thing for you.
The trouble with IPv6 adoption is that you don't see a huge, immediate gain. But in the long run, the costs of staying with IPv4 are going to be higher than the costs of changing to IPv6. We still have enough time to do it fairly painlessly.
I don't buy the 'more addresses' argument. If China wants more IPv4
addresses there are plenty they can take without causing widespread
inconvenience. They could decide to simply advertise BGP routes to net 18 or
one of the other large class As grabbed early on by a US university.
Organizations in China can also simply request the address space from APNIC and get them if the request is within reason, just like American organizations can request address space from ARIN. APNIC, ARIN, LACNIC and RIPE policies regarding IPv4 address assignments are as good as identical. Just because some US organizations were able to get large blocks of address space in the past doesn't mean all US organizations now have all the address space they need.
A few stats: if we squint, 40% of the usable IPv4 address space is unallocated, 20% allocated by the RIRs (= reasonably efficient) and 40% allocated pre-RIR (= mostly rather inefficient).
If I am right in my belief that IP beat OSI through inertia and luck rather
than technical superiority there is a big problem deploying IPv6. It may
well be that inertia wins again and IPv4 plus NAT is good enough for most
uses.
Vendors will have to implement IPv6 because SOME customers require it. (Note that supporting IPv4+IPv6 in applications by using the IPv6 system calls is fairly trivial. The trouble is supporting both the old and new system calls side by side in case you need to run on a system that doesn't have the IPv6 compatible system calls.) Once all hard- and software supports IPv6, the question is no longer "why move to IPv6" but rather "why stick with IPv4".
But we'll really see sparks fly once the Linksyses of this world start to implement automatic 6to4 based on a dynamic IPv4 address (= automatically create, tunnel and give out IPv6 addresses without any need for ISP support) so Gnutella and Kazaa users get to run NATless.