In the future, people might not be able to use IPv4 at all, or all the time, and then you need your website to be available via both IPv4 and IPv6.
We have a /23 of nice IP addresses. There's no ugly NAT or private addressing at our end, so that's not an argument for us to add IPv6 support. They're part of a university's /15 assignment, so it's not realistic to give them up or sell them.
We'll get IPv6 support eventually, I'd guess in 2018-19 or so, but I'm curious if the lack of support causes problems for users in Africa in the mean time.
If a user types in a domain, eventually a root dns responds with an ipv4 A record and resolves to a server the site has configured. Am I missing something?
By supplying v6 you ensure anyone in the world can access it regardless of which IP protocol is deployed by your carrier and how they are or aren't translating from one to the other.
Sounds like this is mostly the case, but it has some issues and any new ISP on v6 would have a hard time because a lot of the big sites would be unreachable.
Of course, this becomes political soon enough. As new countries see mass adoption of the interwebs and find out that they have massively smaller allocations of IP addresses than the existing countries they might feel annoyed and do something about it. Tempers flare, shoes are thrown, everyone gets sulky and unhappy.
All this could be avoided if we just flipped a few switches, wrote some code, and all moved to IPv6. It's such an easy, small thing to do. Why not just do that and not irritate the living buggery out of everyone?
It's about being a good netizen.
So - and this is a genuine question - what is the issue right now with, for example, the BBC not yet having an IPv6 address, as long as they have plans in the pipeline for when v6-only clients need to access them?
Because of the catch-22 problem. The BBC sees that no consumers use IPv6 exclusively, and don't bother supporting IPv6 "yet". ISPs see that "no" content providers support IPv6 at all, and then claim that it doesn't provide any real benefit over CGNAT bullshit.
And so we'll be stuck in this mess for another century at least...
Let's take an example I got from a Cisco presentation, when NAT is used. A map web application may have to download a lot of tiles quickly. It's best done using several parallel TCP connections. Each of those connections must be tracked by the NAT box. At scale and for peak time load, this can become a nightmare and some connections may be dropped --- leading to retries and delays in completing the web page display. With IPv6 this issue doesn't exist.
Supporting IPv6 can help providing a better mobile user experience in this case, and is why a lot of the big companies do support IPv6 already. Bypassing CGNAT helps with latency in a situation as in the above example.
If you make your server available on IPv6, too, then I can use my non-shared IPv6 address to communicate with you which means that the packets don't need to undergo NAT which means a less horrible, faster connection without messy port forwarding requirements.