Predicting IPv6

Predicting the future is easy. Predicting it correctly is incredibly difficult. The only information we have is the past, and for the most part, we try to guess that the future will look kind of like some part of the past.

In the past, new technologies have been adopted according to an S-curve. That is, a slow initial adoption, but then around 5-10% penetration, adoption accelerates. Somewhere around 70-80%, usually, adoption slows down, as the last groups are slow to change. The second table at this New York Times page shows some adoption curves for technologies, where you can see this pattern.

Interestingly, although they are all S-curves, they don't have much symmetry. That is, knowing the early adoption rate doesn't seem to tell you when adoption will accelerate, the rate of the bulk of adoption, or when it will slow down. Math geeks like to take a logistic function to predict the S-curve, and while that does make pretty, symmetric curves, it doesn't look like it's a very good predictor of the future.

One crude way to make a poor prediction of the future of IPv6 adoption is to take the history of adoption and extrapolate it into the future. If you assume that the rate of IPv6 adoption will continue accelerating at the current rate, at least until we hit 50% adoption, you might pretend you have the bottom half of the S-curve.

Google has generously provided statistics over time, measuring (roughly) what percentage of eyeballs viewing them are using IPv6 to do it. The data they use are embedded in that chart. I've imported that data into a spreadsheet (because I'm no math geek) and run some extrapolations (using a 6th order polynomial curve projected forward).

The upshot is that if you throw a dart over your shoulder using previous trajectory, you might think that 50% of eyeballs worldwide will have IPv6 in the first half of 2016. Is that accurate? No, almost certainly not. We just don't know how inaccurate, or how long it will take to get to 70% or 80% or 95% or wherever tapering begins. We also don't know how much content will be available over IPv6, which is the essential part of getting to IPv6-only for any users.

I've made the spreadsheet available, so that you can import updated data and rerun the extrapolations any time you like. And if you're so inclined, you can do the same thing for your favorite country, using Eric Vyncke's data, (replacing “us” is the two-letter code for your country). Here's what I did for the U.S., showing 50% in late 2015.

Lee Howard
16 November 2013