Forum Discussion
@Optimist1 asked:
How accurate are those figures? What does "user" mean? An IPV6 address? If so, the whole 128 bits, or just the routeable 64?
The figures are counts of IP addresses actually observed by APNIC, so APNIC's column heading of "V6 Users (est)" is just shorthand for something more complex. It's not too bad for general public consumption but perhaps a bit misleading.
APNIC's measurements are formally proxies, in science/engineering terminology. I wrote a longish post about them in message 476 of this thread back in April 2018, based on a response I made in the UK IPv6 Council group at LinkedIn:
The question about accuracy is interesting. The accuracy is very dependent on what you're actually trying to measure, since proxies can have a complex relationship to the desired (but hidden) quantity. In engineering terms, the figures have relatively low accuracy because of the very indirect method of measurement and the many things that affect the process. However, their precision can be very high for ISPs that have IPv6 user populations in the millions, since that provides APNIC with very robust statistical populations.
(As a perfect example of very high precision but quite low accuracy, just remember the recent common-mode fault in APNIC measurements that I described in message 750. APNICs figures are very precise for Sky and BT, yet for several days their accuracy and utility as proxies got badly disconnected from the quantity that we're trying to track --- low accuracy.)
In Virgin's case, both the accuracy and the precision are rather low, since the relationship between the proxy measurements and IPv6 user counts is highly uncertain without a public IPv6 release, and the statistical population for their IPv6 counts is pretty low too.
It's worth noting that IPv6 provides end-to-end connectivity from each client device, whereas residential IPv4 users are given a single IPv4 address and all of their devices appear publicly as that single address coming out of their CPE's NAT. Consequently, publicly observed addresses are usually proportional to account numbers for IPv4, but proportional to device numbers for IPv6, which is why APNIC IPv6 counts are typically larger than actual user counts. Proxies are very valuable in the absence of direct data, but one has to tie down what they represent to gain the most benefit from them.
Morgaine.
Related Content
- 6 months ago
- 8 months ago
- 8 months ago