What about IPv6 right here on earth?

If you ask John Curran, CEO of ARIN, and current cheerleader for the IPv6 transition, he'll tell you that the lack of a good v4 to v6 transition strategy was indeed "horribly misconceived", and he said so back in 1994 when the IPng working group first blessed the 3-way merged "simple internet protocol plus" as IPv6. The working group was probably mislead by their memories of the 1981-83 transition of ARPANET from NCP onto TCP/IP v4. Also by the correct theory that a dual transition (v4 to v4+, followed by v4+ to v6) would be twice as expensive and hard to sell as a single transition to v6.

In that earlier transition we were dealing with a much smaller internet (<300 hosts), only 3 protocols (TELNET, FTP, SMTP), a single backbone (BBN), only research organizations as customers, and had a mandated flag day when NCP was turned off and you went off-net if you hadn't converted yet. While the current transition from v4 to v6 lacks all of those characteristics, which is part of what has slowed it, it does share the multiyear transition and immaturity of the new protocol stacks problems we saw last time. I think there will be a flag day eventually too, around 2020, but that will be after the transition already happened (2009-2015?) and after IP traffic on the internet is 99% v6 (2017?).

"The transition is going to be very painful because it makes good commercial sense that way ..."

Yes. But, since there are about to large numbers of users for whom native IPv6 is about to become better, faster, and cheaper than hard to get IPv4 with multiple layers of expensive NAT444 appliances, and the economics are finally going to flip. In the US bad experiences by future 4G smartphone dual-stack-lite (native v6, tunneled + carrier NAT44) customers with v4-only web sites is going to pressure content providers to dual-stack, which is going to pressure ISP's to dual-stack too. This is why the likes of Google, Netflix, Facebook, CNN, and soon Yahoo are already dual-stacked.