Summary

  • On 5 December 2019, AFRINIC-31 considered Draft 1 of “Impact Analysis is Mandatory.” Its authors said earlier analyses had sometimes remained unavailable for more than six months. Their answer was a report for every draft policy proposal, ordinarily within four weeks for a new proposal and two weeks for a revision, with publication and pre-meeting backstops. Supporters saw advance evidence that could improve proposals; critics questioned feasibility, necessity and the micromanagement of staff and chairs.
  • The draft’s central insight was sound but incomplete. Timely analysis could become a procedural entitlement to a common evidentiary record, not a favour distributed by discretion. Yet a report written by AFRINIC would remain one private technical institution’s perspective, not independent truth or permission to govern. The draft did not clearly provide for rebuttal or say what happened when a report was absent. A better rule would make the minimum record visible and contestable, scale depth to likely impact, allow reasoned extensions and defer a consequential consensus call when the record is missing—without giving staff a veto over the proposal’s merits.

A policy room confronted its own information problem

On 5 December 2019 in Luanda, participants at AFRINIC-31 were asked to judge a proposal that sought to put evidence on a clock. The proposal was AFPUB-2019-GEN-005-DRAFT01, version 1.0, submitted one month earlier by Jordi Palet Martinez and Sylvain Aboka Baya. Its title, “Impact Analysis is Mandatory,” stated the remedy before the room had agreed on the diagnosis.

The diagnosis was uncomfortable. The existing wording of section 3.4.1 of the Consolidated Policy Manual allowed the Working Group chairs to request technical, financial, legal or other impact analysis from AFRINIC. The proposal’s authors argued that this discretionary arrangement had not reliably delivered the information needed for judgment. They said some impact analyses had still not been provided after more than six months, including analyses for revised proposals despite staff commitments on the RPD mailing list. The surviving record does not establish every example behind that allegation, so it must remain attributed to the authors.

But their claim framed the immediate question in the room: if participants were expected to assess consequences, how long could the institution hold the assumptions that only it could readily supply?

Draft 1 answered with ordinary deadlines. AFRINIC would have no more than four weeks after publication of a new proposal to provide an Impact Analysis Report. When an existing proposal was revised, the corresponding update would be due within two weeks. For a major review or an unusually complex proposal, the institution could take longer, but only by giving a clear justification and publishing at least a draft report no later than five weeks after the proposal appeared. If that five-week point collided with a public policy meeting, the draft report had to arrive at least ten days before the meeting.

A new version submitted within the final two weeks before a meeting was excepted from that pre-meeting requirement.

Those clocks put two legitimate concerns into direct conflict. One side stressed what advance analysis could do: participants could understand likely effects before the meeting, and authors could correct or refine their text more efficiently. The other side asked whether a universal obligation was practical when many proposals arrived close to a meeting. The minutes also record concerns that the rule could micromanage staff and chairs or require elaborate work where a proposal did not need it. This was not a choice between evidence and ignorance.

It was a dispute over whether a rigid duty would create a usable common record or overload the people expected to assemble it.

Evidence before a consensus call mattered because rough consensus is not simply a count of hands. Its engineering logic asks whether materially supported objections have been identified and addressed. That method depends on the quality of what participants can know. An objection about implementation burden cannot be weighed intelligently if the service dependencies and timing remain hidden. An assurance that a change is harmless is not much stronger if no one has disclosed the assumptions behind it. A meeting can be procedurally orderly and still have a thin evidentiary basis.

The co-chairs did not find consensus. They returned Draft 1 to the list because concerns about necessity and practicality had not been addressed. AFRINIC’s 2020 annual report later listed the proposal as expired. The available record does not show adoption or implementation. The importance of the draft therefore lies neither in a rule that took effect nor in a harm that can be measured after the fact. It lies in the procedural design it attempted: moving access to institutional analysis from an optional request toward an expectation owed to every proposal and every participant asked to judge it.

What the draft would have made visible

The proposal was more than a deadline amendment. It described a report with several distinct fields of inquiry. AFRINIC would explain its interpretation of the proposal and identify possible improvements to the text. It would consider effects on other policies, on the registry itself, on operations and services, and on the membership and wider community. It would set out technical details and an estimated implementation period. It would address the financial impact on AFRINIC and any legal issues it identified.

That range matters because policy language can move through several layers before it reaches a running network. A sentence may look clear in a discussion document but interact unexpectedly with an existing registry rule. A change may require software work, staff procedures, notice to members, security checks or coordination across services. A seemingly narrow obligation may carry an implementation interval that alters when operators can rely on it. Financial costs may fall first on the registry and later appear as service constraints or fees.

Legal uncertainty may affect what staff believe they can do even when the policy room assumes that a consensus call settles the issue.

No outside proposal author necessarily sees all those layers. AFRINIC staff may know dependencies in its registry systems, support operations, internal processes and service calendar that are not apparent from the policy text. They may also be better placed to estimate how long a requested change would take inside the institution. A report can bring that otherwise unevenly distributed knowledge into a common record. It can reveal that a proposed deadline conflicts with technical work, that a provision duplicates another rule, or that an apparently simple instruction has a costly service consequence.

Just as usefully, a short report can state that no material impact has been identified and explain why.

Draft 1 also specified how the report would enter the shared discussion. The final or draft report was to be attached to the relevant proposal and announced on the RPD list within the stated period. Attachment creates a durable association between the proposal version and the analysis written about it. Announcement reduces the chance that only attentive insiders know the document exists. Together, those mechanics would have made it easier for authors and participants to refer to the same version of the same assessment before a meeting.

Version discipline is essential here. A report on an earlier text can mislead when a revised proposal changes the mechanism that created the original concern. The two-week update requirement recognised that the evidentiary record had to follow the text. It did not merely demand a one-time institutional opinion. It sought a view that remained connected to the proposal being judged.

The complex-case extension showed that even the draft’s authors did not treat four weeks as an absolute regardless of subject. They permitted additional time for major review or unusual complexity, provided the delay came with reasons and a draft appeared by a backstop. That is a significant design choice. An unexplained delay leaves authors unable to tell whether work has begun, which questions remain open or whether the report will arrive before the decision. A reasoned extension turns delay into something that can be examined. The draft-report requirement prevents complexity from becoming an indefinite blank space.

Yet the mechanism still concentrated important choices in AFRINIC. Staff would select the assumptions, decide what financial or legal questions deserved emphasis, describe operational effects and choose how uncertainty appeared. The draft called the resulting view neutral, but it did not establish an outside author, an independent reviewer or a procedure for resolving disagreements over framing. Visibility was improved; authority over the first account of consequences remained institutional.

From discretionary assistance to an evidentiary entitlement

The change from “may request” to a report for every draft can be understood as a proposed procedural entitlement. That phrase needs care. It does not mean Draft 1 created an enforceable public right, and it does not mean any author was entitled to a favourable conclusion. It means that access to AFRINIC’s relevant analysis would no longer depend on whether chairs chose to ask for it in a particular case. The institution’s perspective would become a regular part of the process, delivered against visible clocks and shared through identified channels.

Discretion can be useful when demand is varied and staff time is finite. But unexplained discretion also distributes information unevenly. If some proposals receive detailed analysis and others approach a meeting without it, the chairs’ selection affects which authors get a chance to answer institutional concerns in advance. It also affects which participants can test feasibility instead of relying on intuition. A universal minimum report reduces that selection problem. Every proposal receives at least an accountable statement of impact, even where the statement is brief.

The benefit is especially clear for authors. Institutional feedback delivered at or near a meeting can arrive too late for careful revision. An author may hear for the first time that a provision conflicts with a service dependency or imposes work that staff cannot complete on the assumed timetable. The discussion then becomes a rushed choice between defending imperfect text and withdrawing it. Earlier analysis permits a different sequence: disclose concern, inspect the basis, revise if warranted, and let participants judge the revised mechanism rather than a surprise dispute.

Affected operators have a related but distinct interest. Policy-list participation is not a condition of being exposed to registry decisions. People running networks may not attend the meeting or follow every mailing-list thread. Their absence does not remove operational effects. An entitlement to published analysis improves the record available to those who do participate, but it also creates material that can be examined by people outside the immediate room. That is evidence of a more informed private process; it is not consent from everyone affected.

This distinction prevents a useful procedural right from being overstated. The report could make decision-making more legible. It could not transform the policy room into a representative institution for a continent. It could not bind absent operators merely because analysis had been attached and announced. Nor could it turn staff’s view into a legal judgment. The procedural entitlement is modest and important: before a consequential call, the people considering a proposal should be able to see the service operator’s account of likely effects within a known period.

The entitlement also runs to reasons when ordinary timing cannot be met. A clock without an exception can encourage superficial compliance. An exception without reasons can swallow the clock. Draft 1 attempted to join the two by allowing extra time for complexity only with a clear justification and a draft-report backstop. That structure can preserve both candour and predictability. Staff need not pretend a major review is simple, while authors need not accept silence as the price of complexity.

What the draft did not settle was the quality floor for the report. A document could technically touch every listed heading while saying little about sources, assumptions or uncertainty. It could declare no material impact without showing how that conclusion was reached. The text made the existence and timing of the report more regular, but the sealed record does not show requirements for citations, reviewer roles, version history, conflicts or a formal author reply. Access to a document is not yet access to a contestable evidentiary record.

A staff perspective is useful precisely because it is partial

AFRINIC’s perspective has genuine value. It is the service operator’s view of what a proposed rule may require inside the registry. It can describe technical dependencies, administrative effort, member-facing changes, implementation timing, institutional cost and legal concerns that outside authors may not know. Excluding that knowledge would weaken the discussion. Pretending it is independent would weaken the discussion in another way.

Neutrality can be an aspiration: a commitment to analyse the proposal rather than campaign for or against it. Independence is an institutional arrangement. It usually requires some separation between the interests under examination and the person making the assessment, together with rules for appointment, disclosure, conflicts and review. Draft 1 did not create that separation. AFRINIC would assess effects on AFRINIC, its services, its finances and its legal position. Those are subjects on which its knowledge is indispensable, but they are also subjects in which it has an organisational perspective.

Calling the report neutral therefore could not make it conclusive. A cost estimate may rest on a particular implementation choice. A legal concern may assume a scope of authority that an author disputes. A service warning may be entirely sound but omit a lower-cost alternative. A forecast of member impact may reflect the members most visible to staff while missing costs borne by operators who are not present. None of these possibilities implies bad faith. They are ordinary reasons to expose reasoning instead of treating institutional status as proof.

A strong report would distinguish observation from inference. It would identify which existing systems or procedures are implicated, which assumptions connect the proposal to the predicted effect, and how confident the institution is about timing or cost. It would say which proposal version it addresses. It would identify unresolved questions and explain whether outside legal advice had been obtained where legal authority was at issue. Where security or confidential material could not be disclosed, it could give a bounded reason for the omission instead of asking participants to trust an unexplained conclusion.

Authors then need a visible opportunity to respond. They may accept a point and revise the proposal. They may show that the report has read a clause too broadly. They may offer an alternative implementation path or contrary operational evidence. Participants should be able to see both the report and the reply before deciding whether an objection has been answered. If an assumption remains disputed, the record should label it as disputed. The policy room need not invent certainty that the evidence does not provide.

External contributions matter for the same reason. Operators can describe the effect on running networks. Members can explain likely compliance burdens. Technical participants can test a claimed dependency. Counsel can question the legal basis for a proposed action. AFRINIC’s report should organise and disclose its own knowledge, not pre-empt every other kind of evidence. A shared record becomes stronger when the first institutional account is treated as a proposition to test.

This is also why mandatory analysis must not become a staff veto. If staff can prevent consideration simply by withholding the report, the entitlement reverses direction. A safeguard intended to protect authors and affected participants becomes an instrument of agenda control. The answer is not to proceed as though the missing evidence does not matter. The answer is a procedural consequence that preserves the decision until the minimum record exists, while making delay visible and preventing staff from deciding the proposal’s merits by inaction.

The missing rule was what happened next

Draft 1 said the report had to exist, gave time limits and provided publication mechanics. It did not expressly state what followed if the report was late or absent. The text did not say that the proposal became invalid, that discussion was suspended, that a consensus call was barred, that an appeal opened, or that any remedy became available. It also did not establish how authors could challenge the report’s assumptions or seek review of a disputed assessment.

That omission is not a minor drafting detail. Procedural duties derive much of their meaning from the consequences of non-compliance. If a meeting can make the same consequential call without the report, the deadline may become an aspiration. If the proposal automatically dies because staff miss the clock, staff acquire practical control over which proposals survive. If chairs improvise the consequence case by case, discretion returns at the moment it matters most.

The most defensible consequence is deferral of a consequential consensus call until a minimum record and a reasonable response interval are available. Deferral is not a judgment against the proposal. It does not allow staff to reject the text. It preserves the decision because the evidentiary condition for a responsible call has not been met. Discussion can continue, questions can be narrowed, and authors can revise. What should wait is the claim that objections have been adequately addressed when the relevant institutional analysis is still missing.

This consequence must be bounded to avoid abuse. The minimum record should be defined. For a simple drafting correction, it might be a short statement identifying the text, confirming that no material technical, operational, financial, member or legal effect has been found, and showing the basis for that conclusion. For a proposal affecting existing holders or service continuity, the record should be deeper. A public queue should show where the assessment stands. Extensions should give reasons and a revised date. Persistent delay should be visible to participants rather than converted into an invisible stop.

Contestability needs a similarly concrete shape. The report should include its version, sources, central assumptions, unresolved questions and reviewer roles. Authors should have a designated response field linked to the same proposal version. Contrary evidence should be accepted and preserved. Where the dispute concerns a prediction that cannot be resolved in advance, the report should state the uncertainty and identify what would be observed after implementation if the proposal later took effect. This does not guarantee agreement. It makes the basis of disagreement inspectable.

Conflict handling is another gap. Because AFRINIC would be assessing implications for its own finances, operations and legal position, the report should disclose who prepared and reviewed each part. A disclosure does not disqualify staff; it gives participants the context needed to weigh their conclusions. If an outside view was not obtained, the report should not imply that its legal or economic assessment has external validation. Institutional knowledge gains credibility through candour about its boundaries.

The draft’s exception for a new version submitted within two weeks of a meeting reflects a practical limit: authors cannot reasonably expect staff to refresh analysis instantly. But the same logic means participants should not treat the absence of an updated report as harmless. If the revision is material, the call may need to wait. If the change plainly cures a point already analysed, a short update may suffice. The focus should be the information necessary to judge the actual text, not compliance with a ceremonial page count.

The no-consensus outcome demonstrates that these questions were not resolved in December 2019. The minutes record concerns about practicality and necessity, and the chairs returned the proposal to the list. The later expired status shows that Draft 1 did not become an operative rule in the available record. It would therefore be wrong to describe deferral, author response or disclosure requirements as provisions that AFRINIC adopted. They are the safeguards needed to complete the entitlement the draft began to describe.

The strongest case against a rigid clock

Any fair assessment must begin with the best argument for staff discretion. AFRINIC staff held operational, service, financial and implementation knowledge that outside authors could not fully observe. Extracting that knowledge into a reliable report takes time. Work may require input from several functions. A legal issue may not yield a quick answer. An implementation estimate may depend on technical investigation. A sudden cluster of proposals near a meeting can create a queue that no fixed four-week rule can honestly satisfy.

Not every proposal warrants the same depth. A narrow correction to terminology may have no material effect on registry systems or members. Requiring a lengthy analysis in that case can consume time better used on a proposal that changes allocation mechanics or creates service risk. Detailed procedural commands can also constrain chairs and staff when an unusual case calls for judgment. The AFRINIC-31 concern about micromanagement therefore deserves more than dismissal.

There are confidentiality limits too. Staff may encounter security-sensitive details or privileged legal advice that should not be placed on a public list. A rule demanding complete disclosure could harm the very systems it is meant to illuminate. Estimates may also be uncertain. Forcing a firm number or date where none is supportable can create false precision and later distract from the underlying decision.

These concerns explain why a useful entitlement should be proportionate rather than mechanically uniform. They do not justify unreviewable silence. A short no-material-impact report can serve a small proposal. A visible queue can reveal a surge in demand. A reasoned extension can state that several functions must be consulted and give a new target date. A bounded description can identify the category of confidential concern without exposing the protected detail. Uncertainty can be recorded as a range of possibilities or an unresolved question rather than concealed.

Tiering should follow the likely effect of the proposal. A change confined to registry mechanics may need focused technical analysis. A rule concerning future allocation may require broader operational and distributional consideration. A proposal touching existing holders’ transferability, portability, commercial use, fees, recognised control, service continuity or revocation exposure deserves the deepest examination because reliance interests and running networks are already in place. The point of classification is to direct scarce analytical effort toward consequence, not toward the status of the author or the popularity of the idea.

The institution also needs protection from strategic volume. A queue can show when many submissions arrive together and allow sequencing according to meeting dates and likely impact. The ordinary clock remains meaningful because departures are explained. Participants can see whether delay reflects complexity, volume or an unresolved dependency. That information permits criticism where warranted without assuming motive.

Proportionality answers the necessity objection as well. “Mandatory” need not mean that every proposal receives a long paper. It can mean that no proposal reaches decision without an affirmative record: either a developed analysis or a concise, reasoned finding that no material impact has been identified. The duty is to show the work appropriate to the stakes. That is less burdensome than identical treatment and more accountable than a system in which no response may mean either “nothing to report” or “not examined.”

The benign case thus improves the proposal rather than defeating its core insight. Finite capacity supports scaled depth, transparent sequencing and extensions. It does not support a return to unexplained selection. Confidentiality supports bounded reasons, not institutional mystery. Small changes support short reports, not the absence of a record. The four-week clock should be an ordinary expectation around which exceptions are made visible, not a promise that complexity can never disturb.

Costs travel beyond the meeting room

Impact analysis is sometimes treated as administrative housekeeping. In number-resource coordination, its effects can reach much further. Registry rules can influence access to scarce IPv4 resources, transfer processes, service continuity and the compliance work expected of operators. Even where a proposal does not directly change those interests, uncertainty about implementation can alter planning. People must decide whether a transaction, deployment or network change can proceed on the assumptions available to them.

A timely report can identify labour that a proposal will require, dependencies between services, legal uncertainty and an implementation interval. It can expose costs while there is still time to revise the text. A report may show that the intended outcome can be achieved through a narrower mechanism. It may reveal that an obligation falls unevenly on members with different operational circumstances. It may also reassure participants that a feared effect is unlikely, provided the reasoning is visible.

When analysis is absent or late, research costs shift outward. Authors must guess how AFRINIC will interpret their language. Operators must evaluate possible effects without access to the institution’s technical assumptions. Participants at a meeting may have to assess a late concern under time pressure. Delay itself can create reliance risk because people do not know which version or implementation path to plan around. None of this record proves a quantified loss caused by Draft 1, which expired. The mechanism is a reason for precaution, not evidence of a measured outcome.

Institutional framing creates another economic risk. The first detailed assessment often sets the terms of debate. If AFRINIC’s report foregrounds its own cost but says little about burdens on operators, participants may receive an incomplete view of incidence. If it assumes one implementation path, alternatives can appear unrealistic before they have been examined. If its legal position is presented without disclosed assumptions, uncertainty may be mistaken for a prohibition. These are not accusations about staff conduct. They are structural reasons to require rebuttal and multiple forms of evidence.

LARUS’s operator analysis is useful here because it follows governance choices into infrastructure. Hidden assumptions and delayed decisions can become planning costs, service dependencies and business-continuity risks. BTW’s analysis of due process adds the importance of notice, reasons, an opportunity to cure, preservation of the record and review where registry action has serious consequences. This proposal concerned evidence before consensus rather than an appeal from adverse action, so the two situations should not be merged. Still, both show why legibility matters when private administrative choices can affect live systems.

NRS contributes a member-rights perspective that separates legal remedies, collective procedural leverage and continuity planning. That separation helps resist two opposite mistakes. The first is to dismiss evidence rules because they do not create public-law remedies. The second is to describe a private procedural improvement as though it conferred state-like rights. A published report can strengthen an author’s ability to answer, a participant’s ability to object and an operator’s ability to plan without changing AFRINIC’s institutional nature.

Heng Lu’s later Policy Mirror sharpens the fields needed when existing holders may be affected: economic incidence, operational effect, legal authority and protection against retroactive disturbance. That later framework is not proof that Draft 1 contained all those safeguards. The 2019 proposal mentioned membership, community, finance, technical implementation and legal issues, but it did not establish a full affected-principal process. The comparison shows where the proposed report could have gone deeper, especially when policy reaches people who are not active in the room.

Running networks provide the decisive check. Analysis should ask whether a proposal protects the narrow coordination functions that justify a registry’s role: uniqueness, accurate and auditable records, authenticated changes, resistance to duplicate claims and record fraud, reliable contact and security metadata, orderly transfers and continuity. It should also ask whether the proposal reaches beyond those functions into settled operational choices. The farther it reaches, the more important it is to identify who bears cost and risk before calling consensus.