Summary

  • RFC 2050 treated conservation, routability and registration as distinct goals that could conflict; allocation and assignment records established uniqueness and responsibility, not a global route.
  • Its numeric thresholds, appeal path and institutional map describe 1996 practice. RFC 7020 later replaced the document and expressly left actual announcement and advertisement outside the registry system.

The strongest sentence in RFC 2050 is not one of its thresholds. It is the warning inside the routability goal: allocating or assigning IPv4 addresses did not guarantee routability “in any way”. The warning turns what might look like one administrative process into a sequence of separate receipts.

The document began with three objectives. Conservation sought distribution according to operational need and resisted stockpiling. Routability favoured hierarchical allocation so that many addresses could be represented by fewer routing announcements. Registration supplied a public record, protected uniqueness and gave operators someone to contact when technical or security problems arose.

Those aims did not form a neat triangle. RFC 2050 said conservation and routability were often in conflict, while all three could conflict with the interests of an end user or provider. A small allocation might conserve the pool yet create a prefix awkward to carry. A provider-sized aggregate might help the routing table while tying a customer to renumbering. A detailed reassignment record could improve accountability while imposing disclosure and administrative work. The RFC did not offer a universal formula; it required judgment in each case.

That is why its vocabulary matters. A regional registry allocated blocks to an ISP. The ISP could then assign a block to an end enterprise for its own use. An assignment was not permission to subdelegate. A registry entry recorded one of those acts. None of the three was a BGP announcement, a neighbour's import decision or proof that a packet crossed the network.

Registration nevertheless did real work. RFC 2050 required reassignment information to be submitted promptly. It named three purposes: tell operational staff who was using a network number and whom to contact; show that a provider had consumed most of its existing CIDR allocation before asking for more; and support studies of address allocation. The record was both an accountability surface and an input to the next allocation decision.

The evidence could be extensive. A requester might supply subnet masks, host counts, topology, routing protocols and limitations, previous assignments across divisions and subsidiaries, deployment timing, growth estimates and public organisational information. The 1996 document also used immediate and one-year utilisation guidelines and asked for corroboration. These were materials for an administrative judgment. They were not live router configuration, traffic measurements or a deed of ownership.

Routability pulled the system in another direction. RFC 1518 had described an architecture in which topologically related addresses could be aggregated. RFC 2050 therefore encouraged most ISPs to obtain space from an upstream provider and keep CIDR blocks intact. When connectivity ended, customers were encouraged to return the provider's addresses and renumber. The cost of that move was not incidental: it was how routing-table scale was purchased.

Direct registry space remained possible for specified cases such as balanced multihoming or connection to a major exchange. Yet RFC 2050 twice observed that non-provider-based addresses were least likely to be routable. Even documentation from a recognised ISP willing to inject a longer prefix only evidenced one provider's intention. It could not compel every transit network to accept the route.

The operational-guidelines section makes the boundary concrete. Major transit providers might limit prefix size, filter nonaggregated routes or otherwise restrict global advertisements to protect routing capacity. The registry could choose a classless boundary and enter an allocation in its database. It did not operate those transit filters. A valid record and an absent route could therefore both be true.

The same distinction protects against the opposite mistake. Seeing a prefix at one collector does not prove universal propagation. A collector observes the peers that feed it at a particular time. A route may be visible from one region and rejected in another. Even broad visibility does not establish that the origin was authorised, that reverse DNS worked, that a service answered or that the registration criteria remained satisfied.

RFC 2050 linked provider aggregation to return and renumbering. Assignments were described as loans for the duration of connectivity, with sufficient time allowed before reuse. It also said addresses remained valid only while the criteria continued to be met and described possible invalidation when the need disappeared. Those sentences are evidence of the 1996 operating settlement. They are not a present-day instruction to revoke space, a legal conclusion about property or a substitute for the policy and contract that govern a particular resource now.

The date limit begins at the cover. RFC 2050 named three established regional registries: InterNIC, RIPE NCC and APNIC. It set 25 percent immediate and 50 percent one-year utilisation guidelines, an approximately 80 percent reassignment-submission condition, a slow-start model and an appeal chain that could end at IANA. Treating those details as today's global policy would erase the changes the document's successor was written to record.

RFC 7020 replaced RFC 2050 in 2013. It said the Internet Numbers Registry System had changed significantly and omitted policies and operating procedures superseded by ICANN and RIR processes. It updated the institutional account to five RIRs and identified the old final appeal to IANA as no longer appropriate.

Yet RFC 7020 retained the underlying separation. It renamed the goals allocation-pool management, hierarchical allocation and registration accuracy. It also stated that whether addresses are actually announced, and how they are advertised, are operational matters outside the registry system. Replacement did not turn the old document into current law; it preserved a narrower architectural lesson.

The status history is equally precise. RFC 2050 was published as BCP 12 and obsoleted RFC 1466. But the IESG Note did not endorse the policy. It said BCP approval reflected a belief that the text accurately represented current registry practice, expressly withheld recommendation, and promised a December 1997 reevaluation after IRE Working Group discussion. The frozen sources do not establish a distinct outcome for that promised review, so none should be invented.

This is a useful limit on institutional language. Publication status can establish that a document passed a named process. It does not enlarge the claim beyond the process's own note. The IESG attested to descriptive accuracy at a time; it did not certify every trade-off, represent every affected network or guarantee that the rules produced routes.

RFC 2008 later focused on provider-based lending, portability and the possibility that independent networks would filter a more-specific route. That is a different story. RFC 2050's object is the registry decision itself: how scarcity, aggregation and public accountability were weighed before routing actors made their own choices.

The current IANA IPv4 Address Space registry is useful only if read at its proper layer. It records allocation structure today. It cannot retroactively prove that a 1996 request met its plans, that a provider announced a block, that a historical contact answered or that packets reached an endpoint.

RFC 7249 supplies later registry architecture, while Heng Lu's address-book analogy sharpens the operational intuition: maintaining the record is not operating the streets. His reality-layer, running-code and minimum-specification essays are disclosed editorial lenses, not evidence of the RFC authors' intent.

Read layer by layer, the chain is demanding but clear. A registry record can show which allocation or assignment was recorded and which contact was named. The applicable policy and contract show what that record authorised and for how long. Provider configuration can show an intended announcement. Route collectors can show bounded propagation. Probes can show path and delivery from defined vantage points. None may silently stand in for the next.

RFC 2050's lasting achievement was therefore not a percentage. It was an admission of institutional reach. The registry could conserve the pool, arrange addresses to improve the chance of aggregation, and maintain an accountable ledger. It could not press a button that made every autonomous network carry the route.

Sources