I wonder what the risks are for running on "The new version was completely rewritten from the ground up...", but it does seem to have been tested for a few months.
As a developer that does ops stuff as a means to the ends of development - im afraid to admit I don't really understand what I need to do to transition to ipv6. Is it just pointing my domain names at an ipv6 address and updating iptables? Do I have to be concerned with it at an application level?
Generally as a developer you just want to bind to "::" when listening as the equivalent of "0.0.0.0" ie all addresses, or "::1" for localhost (and create an ipv6 socket, but most higher level stacks will do that given an ipv6 adddress). In most stacks that is the only userspace change. The usual setup is that this will also map ipv4 connections over to ipv6. That is not always true, eg OpenBSD will never bind to both kinds of address, you have to listen on two sockets, and you can force that behaviour on other systems. This should work so long as the host you are on does not have ipv6 completely disabled in the kernel, even if there is in fact no ipv6 routing.
Most things "just work", though there are minor fixes and things to know, eg the standard notation for ipv6 literals in http is http://[::1]/ as the : might be a port specifier. I think the only thing I used recently that had no idea about ipv6 was Twisted, oh and some Java applications.
It really depends on how deep in the stack you're working. Apache and Nginx already have support for IPv6. If you're using something like node.js directly, you'll need to bind to an IPv6 port as well.
As a developer that does ops stuff as a means to the ends of development - im afraid to admit I don't really understand what I need to do to transition to ipv6. Is it just pointing my domain names at an ipv6 address and updating iptables? Do I have to be concerned with it at an application level?