I did not yet read the HTTP/2 draft but I've read the SPDY draft about 2.5 years ago. So I might be wrong here with my assumptions.

We're now about 17 years into the IPv6 migration and still we do not have widespread IPv6 adoption on the client side. That's mostly an issue of CPE device rotation at the customer side and updating provisioning solutions. For the server side we're more or less done. Operating system support is there, commercial firewall and router vendors picked it up in the last ten years, datacenter providers are ready, so we're waiting for the users to put some pressure on providing a service via IPv6.

Looking at HTTP/2 it's also a different protocol. The gap might be somewhat closer than the one between IPv4 and IPv6, but it's still huge. Now I'd bet that in the case of HTTP/2 we'll also see a really slow adoption, but this time it's not the client side that's holding back the adoption. For HTTP/2 I've no doubts about a fast adoption on the client side. Google and Mozilla are nowadays providing some kind of continues delivery of new features to the end user on a monthly basis (surprisingly that also works for mobile devices!). So the web browser near you will soon have a HTTP/2 implementation. Even Microsoft is switching to a new development model of rolling updates with Windows 10 and the Internet Explorer successor. But looking at the server side I doubt we'll have widespread HTTP/2 support within the next 5 years. Maybe in 10. But I doubt even that.

With all those reverse proxies, interception devices for header enrichment at mobile carriers, application servers, self written HTTP implementations, load balancers and so on I doubt we'll have a fast and smooth migration ahead. But maybe we're all lucky and I'm wrong. I'd really love to be wrong here.

Maybe we can provide working IPv4-IPv6 dual-stack setups for everyone in the meantime.