Also Facebook (was: Re: Dual stack IPv6 for IPv4 depletion)

Ricky Beam jfbeam at gmail.com
Sat Jul 11 00:58:22 UTC 2015


On Fri, 10 Jul 2015 06:14:16 -0400, John Curran <jcurran at arin.net> wrote:
> If there are “holes” in the methodology, then they are quite consistent  
> holes...

They are mere statistics. They say only what they say without any measured  
margin of error.

For Google, their numbers are collected via javascript embedded in search  
results. Anything that prevents that JS from running to completion  
(noscript, error, nagivating away, ...) is a lost count. Anyone not using  
Google search won't be counted. (that's a non-zero number, btw. But  
likewise, is difficult to prove.) So, there's ways to be missed in their  
numbers. OTOH, those being counted could, potentially, be counted more  
than once depending on how much of the address they correlate. (privacy  
extensions rotate the address)

I don't know how Facebook is collecting stats. But I suspect they have a  
wider sampling base due simply to the fact almost every web page on the  
internet has some content pulled from facebook -- eg. comment engine,  
authentication engine, or "post on facebook" pingback button. How many  
sites do you visit per day that pull nothing at all from facebook?

The only people who can give solid numbers w.r.t. IPv6 usage are the ISPs  
themselves. And we cannot trust them because it's a marketing statistic,  
if they release anything at all. (been there, watched marketing/PR ignore  
my numbers and make up their own.)



More information about the NANOG mailing list