This post does not reflect the views on current, past or future employers. The opinions in this article are my own.
I have worked on the next generation Internet Protocol, IPv6 for near 25 years. I have worked on implementations, standardisation and deployment of pretty much every aspect of IPv6. It’s not an understatement that I have spent most of my working career on the protocol development and the transition of the Internet from IPv4 to IPv6.
In the Internet community we have been adamant that the transition to IPv6 was inevitable, and that there was no plan B. What if we are wrong? IPv6 deployment graphs seems to have flattened out. Deployment stagnated. If an end-user is IPv6 enabled, about half the traffic uses IPv6. Many Internet providers and the biggest content providers have all enabled IPv6. But there is little to no deployment in Enterprises, and there is no evidence of an increased uptake in deployments from the laggards.
Metcalfe’s law states that the value of a network is proportional to the square of the connected users. Nearly all nodes are connected to the IPv4 Internet and less than half are connected to the IPv6 Internet. Those who have deployed IPv6, still need to continue access to the IPv4 Internet. Until the IPv6 Internet has the same value as the IPv4 Internet; would it not make most sense to sit on the fence and wait instead of deploying IPv6 now? In fact, there is an increasing concern in the Internet community about the cost of maintaining a dual-stack network. While the IPv6 proponents have pushed the envelope in mechanisms to make more and more of the parts (applications, host stacks, access network) IPv6 only, access to the IPv4 Internet is still required. Compared to just keeping the network IPv4 only, it’s hard to justify the added complexity. The IPv6 litmus test: If the IPv6 Internet access is broken. Would a user notice?What if the access to the IPv4 Internet was broken?
There are values and politics embedded in any protocol. IPv6 tries to network like it is 1991. With every node on the network having global addresses re-establishing transparent end to end connectivity. While that is a laudable goal the Internet has managed quite well without it. Exposing nodes to global addressing does not come without a cost though. IP addresses are ephemeral and can change at the whim of the service provider or as location and network attachment changes. 30 years ago an IP address was envisioned to be used as a permanent identifier, but that did not turn out to happen. The lack of end to end transparency may have led to today’s centralized Internet. Although there are many other factors that have led to the centralisation of services.
That premise that every host should be accessible by every other host on the network, is of course laudable. But it still need to be secured. It is not obvious that a solution with end to end global addressing is easier to secure than a network of networks connecting via addressing gateways (NATs).
Conclusion Link to heading
If the rational choice is still to sit on the fence and not deploy IPv6, then will ever IPv6 deployment happen? The trough of disillusionment cannot have lasted for 30 years?
If we accept that IPv6 is a failure, what do we do then, is there a plan B? Continuing to evolve the current IPv4 Internet protocol seems to be the obvious answer. Move away from the use of well-known ports to increase the mapping space for end-point dependent NATs. Perhaps implement a door knocking protocol, so services can be hosted behind NATs…? The promise that IPv6 is just around the corner has blocked evolutions in the IPv4 Internet. Thinking of IPv6 as a failure may help in allowing new ideas in the IPv4 Internet to flourish.
What will happen with IPv6? Link to heading
Just like MPLS, IPv6 has found a niche in service provider networks through mechanisms like Segment Routing for IPv6 (although that could also have been implemented in IPv4), Carrier Grade NAT bypass in Service provider access networks (if we think carrying flow state is too costly and not required anyway), and in IOT deployments. IPv6 will probably not go away for quite some time, but it’s unclear if it will add value outside of these niches.
Government regulation is still a wildcard. Various governments have instituted regulations regarding IPv6 deployment. Last time around this track that was not particularly successful.