Summary

  • World IPv6 Day tested IPv6 on participating websites for 24 hours on 8 June 2011. It did not ask every access network or device to change protocol.
  • World IPv6 Launch, beginning 6 June 2012, assigned different jobs to websites, wireline ISPs, home-router makers and CDNs. Its 1% ISP goal applied to a bounded group of residential subscribers who visited participating websites.
  • Measurements show a visible but partial residue after the test. DNS records, working connections, user capability, protocol preference and traffic share are different observations; none alone proves that the Internet switched to IPv6.

The number that fell, then stayed higher

One year after World IPv6 Day, an RFC measured a trace of the event in DNS. Among 10,000 popular sites, only 116 had an IPv6 address record for the familiar www name on the day before the test. During the event, that count rose to 374 for three consecutive measurement runs. Afterward it dropped sharply, but settled around 160 rather than returning to 116.

The shape of that line captures the history better than the word “switch.” The event was temporary by design, and many sites turned IPv6 off again. Some did not. The Internet Society’s 2011 report said more than 1,000 organizations had joined the effort, while nearly 400 enabled their main services for the 24-hour period. Those figures describe different things: the first is organizations, the second is participating services. Neither means that a thousand websites or the entire Internet changed over.

World IPv6 Day took place on 8 June 2011. Participating sites made their services available over IPv6 alongside IPv4, exposing the new path to ordinary visits rather than a lab-only test. The scope was intentionally limited. The later Launch FAQ says the 2011 event involved websites only. That boundary made the trial useful: large content providers could exercise real traffic, while users whose networks were not ready could continue reaching the same site over IPv4.

What a successful test could establish

The event addressed a coordination problem. A website had little reason to operate IPv6 if few customers could reach it; an access provider had little reason to enable customers if there was no content worth reaching. Google’s January 2011 announcement described the goal as finding unexpected issues before wider deployment. It estimated that 99.95% of users would be unaffected, while warning that misconfigured home-network devices could cause trouble in rare cases. That was a forecast, not a measured account of every user.

Later company accounts were reassuring but local. Microsoft said its web properties saw very few affected users and reported no incident on its own services. The company also described three aims: verify website support, carry substantial web traffic over IPv6, and learn what engineering problems appeared at scale. That is evidence about Microsoft’s experience and intent. It is not a census of all paths or households.

The more revealing record came from measurements that kept going after the event. RFC 6948 used an Alexa Top 10,000 list from 1 June 2011 and ran tests every three hours through 11 July. The authors attempted DNS lookups and TCP connections to port 80 over IPv4 and IPv6. Most of the reported results came from one measurement point at Ericsson Research Finland, near Helsinki, though the work also ran at two other sites.

The DNS series counted different things depending on which name was queried. Before the day, 274 sites had IPv6 in DNS for at least one of several tested hostnames, but just 116 used the ordinary www name. At the event peak, 491 of the 10,000 sites had both A and AAAA records across the tested names; after the event, more than 300 still did. For the common www host, the peak was 374, followed by roughly 160. That is a persistent but incomplete change in a selected list, not the fraction of all websites or people using IPv6.

A record in DNS did not guarantee a successful connection. The RFC’s TCP checks found IPv6 names that did not answer on port 80. Some names were not meant to host web pages; others had configuration problems, and the authors said their sample and measurement point limited generalization. That distinction matters: publishing an address is one operational step, while a working end-to-end service is another.

A separate RIPE Labs analysis asked a different question using daily measurements of Alexa-listed sites. It reported that IPv6-reachable sites rose from about 0.4% to 0.7% after the day, with a higher event-day spike. It also observed that networks announcing IPv6 prefixes had grown more quickly in the weeks before the event. The author suggested that operators may have treated the date as a deadline, while presenting that explanation as tentative. This was a second view, not a replacement for the RFC’s DNS and TCP series.

Some participants kept IPv6; others paused

The trial did not leave every service in the same state. Microsoft reported that Bing and Microsoft.com disabled IPv6 at the end of the 24-hour period while the company continued work toward permanent support. Xbox web properties kept IPv6 enabled indefinitely because the event work had provided much of the infrastructure they needed. One company’s different products therefore made different choices after the same test.

That was not a contradiction in the event’s design. A time-boxed test can expose a path, surface broken configurations and lower uncertainty without requiring every participant to declare permanent readiness on the same day. Its durable effects were distributed: some records disappeared, others remained, and operators had new observations to plan from.

Launch changed who had to act

On 6 June 2012, the Internet Society coordinated World IPv6 Launch. The new program moved beyond the websites that had anchored the earlier test. Participating web firms committed to keep IPv6 on their main sites. ISPs were to make it available automatically to a significant portion of their customers. Home-router manufacturers committed to IPv6 by default across product ranges. Akamai and Limelight said they would enable customers through IPv6 across their delivery infrastructure.

The ISP target is easy to misstate. The launch announcement asked each participating wireline ISP to reach at least 1% of its residential subscribers who visited participating websites using IPv6. It did not say that 1% of all global Internet users, all of an ISP’s subscribers, or all Internet traffic had switched. The denominator was deliberately bounded by a type of subscriber and a set of participating destinations. The modest threshold also acknowledged that some households would need new routers or updated software before they could use IPv6.

The event measurement page makes the denominator issue visible. Akamai counted IPv6 requests on selected dual-stack properties. Facebook sampled page loads and measured unique IPv6 users within the ISP cohorts it observed. Google sampled visits to its properties and compared requests to IPv4-only and dual-stack endpoints. APNIC recruited client-side tests through web advertising. These instruments differ in population and unit: requests, users, network capability or protocol preference.

The page listed operator results when at least two website sources were available and their simple average exceeded 0.1%; it also warned that some networks lacked enough data to appear.

That evidence could support a narrow question about participating access networks and websites. It could not produce one neutral percentage called “the IPv6 Internet.” Google’s June 2012 account said the new service would run in parallel with IPv4 and that complete transition would take years. The continued presence of both protocols was the operating plan, not a failed switch.

The achievement was a change in the assignment

World IPv6 Day tested whether a major website could expose a second protocol path without disturbing the first. Launch tried to make that second path part of normal service by aligning content, access and equipment commitments. The first event changed what operators could observe; the second changed what participating organizations said they would keep enabled.

The distinction is smaller than a global migration and more important than a publicity date. A deadline gave organizations a shared point to prepare, compare results and coordinate work that depended on one another. The record that remains is mixed: some sites withdrew their event-only records, some left service running, network deployment continued, and measurement methods described different slices of use. IPv6 became more available through repeated operational decisions. No single day turned the Internet off IPv4.

Sources and records