Summary

  • RFC 3177 recommended a /48 for an end site in the general case. RIR policies adopted that advice, but the policy debate reopened in 2005 as the range of end-site needs became harder to fit to a single default.
  • RFC 6177 did not say “give homes only a /64.” It retired the one-size-fits-all /48 recommendation, warned that fixed prefix boundaries might become hard-coded, and kept the expectation that sites receive enough space for multiple subnets and growth over years.

A clean answer to a messy boundary

IPv6 address planning has two distinct edges. One separates the large allocations held by registries and providers; another is the boundary between a provider's public network and the sites it serves. RFC 3177 addressed the second. In September 2001, the IAB and IESG recommended /48 for an end site “in the general case,” /64 when exactly one subnet was known to be needed, and /128 only when exactly one device would connect. The document was Informational: it recorded guidance to the address registries, not an Internet Standard binding every ISP.

The /48 had an appealing operational logic. It left a site sixteen bits to number subnets, gave room to reorganize a campus or household network without returning for another block, and could reduce repeated review of customers' projected need. A common boundary also made it easier to explain plans and to keep assignments legible. RFC 3177 connected that flexibility to less renumbering and less pressure to use IPv6-to-IPv6 translation merely to conserve addresses.

Those benefits depended on treating “site” as a useful unit. A university, a mobile network, a home with one LAN, and a small office do not necessarily have the same topology or growth path. When a default becomes a fixed table of /48, /64 and /128, it may look less like an adaptable example and more like a rule software and operating procedures are expected to recognize.

Policy moved before the RFC did

The sequence is important. RFC 6177 says the regional Internet registries adopted policies consistent with RFC 3177 in 2002. In 2005, APNIC, ARIN and RIPE communities began revisiting end-site assignment policy; proposals considered /56 for smaller sites and left more of the exact choice to an ISP or local registry. RIPE meeting minutes record debate about whether regional policy should prescribe a provider's assignment size. The records show a policy process, not a census of what providers were actually assigning.

The later IETF document arrived after that operational reconsideration had begun. RFC 6177, published in March 2011 as Best Current Practice 157, states that exact assignment size belongs to the operational community. It explains why: end sites vary, and defining only a few prefix boundaries risks hard-coding those lengths into implementations, echoing classful addressing. “Classful” here is a warning about inflexible recognition of prefix lengths, not a claim that CIDR ceased to work.

RFC 6177 did not discard the older concern about growth. A site should be able to obtain enough space for planned needs over years rather than months, with multiple subnets generally possible. It rejects both extremes: a compulsory /48 for every site and a default so small that customers must continually justify incremental requests. The RIR proposals' /56 examples are policy choices, not proof that every ISP changed its retail practice at once.

A recommendation with a smaller jurisdiction

This is the useful historical turn: the standards bodies kept an architectural voice while acknowledging a limit to that voice. RFC 3177 offered an elegant common default at the public/private edge. After the registries had to turn that advice into assignment rules, the operational communities revised policy first. RFC 6177 then replaced the universal-default recommendation with a sufficiency principle and more room for local choice.

That change does not tell a present-day household whether it needs a /56, a /48 or something else. Nor does a prefix assignment by itself prove a route was advertised, a subnet configured, or a service reachable. It documents how IPv6's plentiful space was translated into policy: not by one perfect size, but by distinguishing architectural guidance from the operational decision about which sites receive what.

Sources

Evidence boundary

The RFCs and regional policy records establish recommendations, policy text and recorded process. They do not establish a universal ISP assignment, actual household subnet counts, depletion avoided, or customer outcomes caused by either prefix length. A prefix is not automatically a route, a configured subnet, reverse-DNS delegation or working service.