"the worse the better"
I think the next big step for ISP's is to provide IPv6-only native service to their clients, and add IPv6-to-IPv4 gateways to allow access to IPv4-only internet.
Isn't that pretty much how DS-lite works?
Edit: for my money, NAT64 + DNS64 would get us 90% of the way there. OS support for pseudo-IPv4 would really fix this 100%: if the application wants to talk IPv4, the OS would transparently translate it into an IPv6 address in the 64:ff9b::/96 address prefix.
Personally, DS-Lite seems cleaner but I don't think it's worth arguing about.
From slide 8:
> Mobile networks don’t use DHCP, so no way to setup MAP or DS-lite without some heavy lifting in protocols and standards
Why can't the handset itself do the DS-lite IPv4 NAT/encapsulation, or why is 464XLAT easier to do on handset than DS-lite?
edit: found this gold-nugget of a slide: http://i.imgur.com/VhMSA2p.png ... ugh, just kinda weird/sad that this is still such an open problem.
Yeah, I've made exactly the same argument. As you found, there's a lot of history here. As I remember it, T-Mobile was originally pitching NAT64 because it required no support on the phone at all and DS-Lite would require the phone to do encapsulation. People pointed out that DS-Lite supports v4-only apps and NAT64 does not, but it seems like T-Mobile chose not to hear that argument. Then, having already committed to the NAT64 route, they added a stateless NAT46 agent on the phone to fix v4-only apps.
I think that agreeing on a standard is more valuable than continually iterating towards perfection, so I wish the industry had just declared DS-Lite "good enough" and stopped, but now we have this menagerie of transition schemes.