Summary

  • RFC 1380 separated immediate operating relief, short-term CIDR work, a mid-term answer to address exhaustion and long-term Internet-layer research. It said all four phases had to start immediately, not wait in a queue.
  • The memo called CIDR a way to buy time under the existing address scheme. It also required policy, protocol, implementation and deployment work, while admitting that interim measures could divert resources from the long-term solution.
  • Its most durable lesson is evidentiary: a recommendation, a slowing indicator, a transition plan, a selected architecture and a deployed result are different records owned by different actors.

The crisis had three clocks

In November 1992, RFC 1380 described three growth problems that were easy to blend together. Class B network numbers were being consumed. Routing information was expanding beyond the comfortable capacity of routers and operators. Farther out, the 32-bit address space itself might become inadequate.

Those problems shared a cause—growth—but not a deadline or a remedy. Giving a medium-sized site several Class C networks could conserve Class B numbers while adding routes. Buying larger routers could absorb more entries without changing the allocation structure that created them. A larger address space could postpone numerical exhaustion while reproducing routing-table overload on a larger field.

The memo also divided routing pressure into machine and human limits. Memory, processing and update bandwidth constrained routers. Configuration and traffic monitoring constrained operators. The first category could invite a hardware purchase. The second required someone to understand whether policy had been expressed correctly across a growing number of routes. A device count and a labour burden were related, but neither measured the other.

RFC 1380 reported that network-number assignments and entries in the Merit NSFNET routing database had roughly doubled every twelve months. That was serious evidence from a named period and observation system. It was not a census of every router, nor proof that a forecast would mature on schedule. The number established pressure; the decision still needed a horizon.

ROAD was a temporary place to deliberate

The Routing and Addressing group was not a standing standards body with a permanent mandate. RFC 1380 describes ROAD as a one-time special group, formed after the Santa Fe IETF and asked to report at the San Diego meeting. It met face to face, discussed by electronic mail and fed proposals into BOFs, plenary discussion and prospective working groups.

That structure matters. A temporary group could gather urgency, compare options and recommend work. It did not automatically own every later standard, allocation policy, router implementation or local migration. The memo itself is labelled Informational and calls itself a preliminary report of IESG deliberations. It records a decision process in motion, not an Internet Standard and not an execution log.

ROAD converged on CIDR as a near-term direction for route aggregation. It did not produce a final recommendation for one larger-address architecture at San Diego. The asymmetry was deliberate: one problem had a sufficiently bounded intervention; the other still lacked enough transition and impact evidence.

One cure could not arrive on every clock

RFC 1380 rejected the comforting idea that one design could solve near-term routing pressure, address exhaustion and advanced Internet-layer functions before the nearest limits arrived. A new Internet layer implied changes to a large installed base. The technical questions were not yet mature, and selection, implementation and deployment would take time.

The IESG therefore described at least two major stages and named four working horizons: immediate, short-term, mid-term and long-term. The labels sound sequential. The instruction was the opposite. All phases should start immediately; the community could not afford to pursue them consecutively.

That sentence changes the management model. “Immediate” no longer means that every person moves onto the nearest fire. It means that each horizon receives an accountable start, even when its finish date differs. Otherwise the urgent layer consumes the calendar repeatedly, and the long-term layer begins only after the temporary system has become an installed base of its own.

Immediate action was not resolution

The immediate list included more conservative address assignment, alignment between allocation and aggregation, reclaiming unused Class B numbers, stronger routers at key points and topology engineering. None required a new protocol. RFC 1380 explicitly said none solved the underlying problems; they could slow their onset.

That distinction protects the record. A stricter allocation decision proves that a policy changed. A recovered number proves that a specific resource returned to the pool. A router upgrade proves that some equipment capacity changed. A smaller table at one observation point may show relief. None alone proves that address exhaustion, operator workload or transition risk has been resolved.

It also exposes distributional questions. Conservative assignment transfers proof work to applicants and judgment to registries. Topology engineering may trade resilience or path quality for a smaller state burden. More powerful routers move cost to operators. An “immediate” action can be reversible in protocol terms while creating durable administrative expectations.

CIDR bought time through a chain of owners

RFC 1338 framed supernetting as a short-term bridge. It aimed to slow Class B consumption and routing-table growth while a long-term answer was developed. Its own expectation of at least three years was a planning assumption, not a warranty.

The intervention was not one switch. Addresses had to be assigned along aggregatable boundaries. Inter-domain protocols had to carry arbitrary network-and-mask pairs. Routers had to implement and deploy the semantics. Non-CIDR domains and interior protocols needed workable boundaries. Multihoming and provider changes could preserve more-specific routes. Deploying allocation policy without capable routing could even accelerate table growth temporarily.

RFC 1380 accordingly distributed the programme: an operational addressing plan, BGP work, possible IDRP work and explicit deployment planning. It called CIDR a way to address routing-table explosion under the existing address scheme, make Class B pressure less important and buy time. It did not call the purchase free.

RFC 1519 later revised the CIDR specification and retained the bridge language. A revised document proves maturation of the plan. Actual relief still depends on implementation, participation, topology, observation point and time.

The bridge could consume the destination

RFC 1380 made two long-term costs unusually visible. First, replacing or extending the Internet layer would be traumatic for vendors, operators and users. Short- and mid-term work therefore had to minimize that trauma or preserve a smooth transition path. A locally successful bridge could still be a strategic failure if it made exit harder.

Second, interim development and deployment would divert resources from other projects, including the long-term solution. This is not a footnote about staffing. It is a claim about opportunity cost. The same engineers, testbeds, operator attention and change windows cannot be assigned twice merely because budgets describe the work under different headings.

Time bought should therefore be recorded alongside time spent. If a short-term measure adds eighteen months of capacity but absorbs two years of migration work, its headline benefit and programme effect point in different directions. RFC 1380 did not supply those later measurements. It did insist that the diversion be considered before choosing the measure.

Criteria came before a bigger address decision

The IESG declined to select a larger-address proposal when the transition evidence was incomplete. RFC 1380 named operational infrastructure, effects on existing protocols, new routing, address assignment, performance, ownership of change control, management, security and training costs. Appendix B asked each proposal to respond point by point, group changes by affected devices and functions, and disclose implementation experience.

This was not indecision disguised as process. Urgency set a deadline, but urgency did not turn an attractive proposal into evidence about migration. A call for proposals, an Internet-Draft, public review, a presentation, an IESG recommendation and an IAB decision remained separate artifacts. The timetable described who should produce what. It did not certify the result in advance.

The later recommendation remained a separate act

The decision process did not end inside RFC 1380. RFC 1719 later recorded that an IPDecide BOF had highlighted the absence of a firm direction. It placed responsibility for developing an IPng recommendation with the IESG, required an open and pre-announced process, and joined urgency estimates to allocation rates, policy effects, expected CIDR savings and the time needed to develop, field and migrate a new IP.

In January 1995, RFC 1752 documented the later recommendation. It selected a revised SIPP proposal as the basis for IPng, separated protocol, autoconfiguration, transition and coexistence work, and recorded that version number 6 would be called IPv6.

That later choice does not turn the 1992 memo into a hidden IPv6 decision. It proves the opposite: long-term selection needed another process, a broader evidence packet and a new accountable recommendation. Nor does the recommendation prove universal migration or the retirement of IPv4. Selection, specification, deployment and outcome kept their own clocks.

Sources and limits

This account uses the official texts of RFC 1338, RFC 1380, RFC 1519, RFC 1719 and RFC 1752. They establish contemporary problem statements, recommendations, criteria and document succession. They do not establish the state of a named current network, the accuracy of every forecast, completion of every milestone, a particular implementation, an authorized local change or a measured operational result.

The bounded conclusion is narrower. In 1992, the Internet's stewards documented that urgent relief and long-term replacement could not be treated as one project or as consecutive projects. A temporary measure needed a benefit record, a resource ledger and an exit path. Without all three, the system could count the time it bought while losing sight of the future it was spending.