Summary

  • At APNIC 14 in September 2002, Philip Smith proposed reserving ordinary-looking IPv6 addresses for documentation. The meeting backed the principle and left the exact prefix for later coordination with other RIRs.
  • APNIC later allocated 2001:db8::/32. Smith, Geoff Huston and Anne Lord wrote RFC 3849, which put the block on the page and told operators how to keep it out of real routing.

A fictional router deserves fictional neighbours

Try writing a convincing network tutorial using only addresses with special local behavior. A reader learning a normal global unicast deployment sees an example that behaves unlike the network being taught. Borrow a real address instead, and the example may point at somebody else's equipment when copied. In the APNIC 14 policy session, Philip Smith of Cisco asked for a third option: a piece of IPv6 space reserved for documentation.

The room recognized the practical itch. IPv4 and ASNs already had numbers suitable for examples; IPv6 did not. Participants wanted one prefix usable globally and examples that looked like ordinary addresses, rather than link-local or site-local ones. They debated where such a block could come from. Smith had brought a principle, not an exact hexadecimal answer. The chair found consensus on that principle and sent APNIC to consult the other RIRs about the range. The APNIC member meeting recorded the same boundary. It takes restraint to approve a solution before choosing the magic number.

Two years later, the number had a name: 2001:db8::/32. RFC 3849, authored by Geoff Huston, Anne Lord and Smith, says APNIC allocated a conventional-size unicast block so documentation could resemble actual deployments. The RFC recommends putting the prefix on non-routable lists and, where filters are used, adding it there too. No end party receives it. That distinction makes a manual's topology realistic while keeping its sample addresses out of the working Internet.

The IANA registry still names the block as documentation space. Smith's APNIC speaker biography describes years spent designing and teaching ISP networks. The small 2002 proposal fits that working life: a trainer notices that the numbers on a slide can escape the slide. He helped give the examples a proper address, and operators a clear reason to refuse it as a route.