In a recent unscientific poll, I asked members of an IPv6 group on Facebook:
“In what year will 10% (or more) of new ISP customers be IPv6-only (or IPv4-degraded)?”
39% of respondents said IPv4 would be degraded or unavailable for many users in 2014, and 89% said that would be true by 2017. There was a fairly even distribution among the years through 2017.
| Year | Percentage of Respondents Answering |
| 2014 | 39% |
| 2015 | 14% |
| 2016 | 19% |
| 2017 | 17% |
| 2018 | 3% |
| 2019 | 3% |
| 2020 | 0% |
| 2021 | 0% |
| 2022 | 0% |
| 2023 | 6% |
The question is interesting because it doesn’t assume that everything will be IPv6-enabled, just that IPv4 will be not as good, if it’s there at all. I didn’t ask “How degraded?” but opinions could range from home NAT, carrier-grade NAT, or some translation technology, to simply not being available. It could also mean “not as fast as IPv6.” Geoff Huston of APNIC has been running IPv6 experiments using Google ads, and as he showed at the recent RIPE-67 conference, IPv6 is faster than IPv4, most of the time.
From Geoff Huston’s RIPE67 presentation, slide 35
What does that mean?
I interpret it to mean that if you host a web site, you probably want to make sure it supports IPv6 before users suffer on IPv4. Some people who are well-informed about IPv6 think that’s pretty soon; indeed, for most people, IPv6 already provides a slight performance edge over IPv4.