Summary

  • Draft 7, submitted on 17 May 2021, paired a mandatory, regularly checked abuse-c contact with an express statement that defining abuse and escalating it under applicable procedures and law were outside AFRINIC’s scope.
  • That boundary was not fully reflected in every clause: validation still referred to measures and responses, while an allegedly incorrect or missing response could trigger revalidation.
  • A private registry can test whether a declared channel is reachable and keep its records accurate. It cannot use that test to decide the truth of a complaint, legal responsibility, adequacy of a response or sufficiency of remediation.
  • A narrower design would preserve the economic benefits of a uniform role mailbox, provide notice and at least 30 days to cure, and prevent contact failure alone from disturbing resource, governance or operational continuity.

A useful contact rule meets its institutional limit

On 17 May 2021, AFRINIC received version 7 of a proposal identified as AFPUB-2018-GEN-001-DRAFT07, authored by Jordi Palet Martinez and directed at section 8.0 of the Consolidated Policy Manual. The text contained a seam that should determine how the proposal is read. One passage said AFRINIC did not define abuse because abuse lay outside its actions; each participant would define and escalate it under its own procedures and local law, beyond AFRINIC’s scope. Yet the validation procedure retained objectives concerning whether a resource holder took measures and provided a response. It also allowed alleged gaming, an incorrect response or no response to be reported to AFRINIC, potentially triggering revalidation.

The first half of that arrangement addressed a real coordination failure. An abuse report is useless if the address is stale, hidden or attached to a person who left years ago. A mandatory role contact can give reporters a predictable destination. A delivery check can expose an address that no longer works. Regular maintenance can improve the quality of WHOIS data, lower the time spent hunting through websites and personal contacts, and make reporting across organisational or regional boundaries less erratic.

The second half raised a different question. Once a complaint about an “incorrect” response reaches the registry, what exactly is being checked? A message can be delivered or rejected. A mailbox can be monitored or abandoned. Those are observable contact facts. But whether the conduct described in a report is abusive, whether the recipient is legally responsible, whether a reply is adequate and whether any remedy is sufficient are contested judgments. They depend on definitions, evidence, jurisdiction and fair procedure. AFRINIC is not the institution empowered to make those judgments.

This is not an argument for indifference to abuse. It is an argument for assigning different functions to institutions equipped to perform them. AFRINIC is a private technical bookkeeper and coordinator. Its valuable common work includes maintaining unique-number records, accurate contact information, objective proof of control, interoperable updates, security assertions, dispute metadata and operational continuity. That role supports the network precisely because it is narrow. The registry can maintain the address book. It cannot turn the address book into a forum for deciding the case inside an envelope.

Draft 7’s operative architecture required an abuse-c attribute in inetnum, inet6num and aut-num objects. The attribute would point to a person or role object with a valid abuse-mailbox that was monitored and actively managed. Access was intended to be unrestricted through WHOIS, application interfaces and future mechanisms. A child object could inherit its parent’s contact or use its own. These provisions made the proposed channel discoverable and attached it to the records through which network resources are coordinated.

The expected mailbox had to be capable of receiving reports with supporting material such as logs, examples and headers. Draft 7 did not impose one report form, and it expected the messages to require recipient intervention. The validation objectives went further than pure deliverability: they included confirming that the holder had read the relevant procedure or policy, regularly monitored the mailbox, took measures and provided a response to reports.

Timing gave the mechanism practical force. The first validation window was no more than 15 days. After failure, AFRINIC could turn to other registered contacts and allow a second period of no more than 15 days. Validation was specified at creation or update, at least every six months, and whenever AFRINIC considered it appropriate. Reports of gaming, an incorrect response or no response could initiate another check of the abuse-c contact.

The proposal also formalised section 8.7. Once per year, AFRINIC could change validation periods or frequency according to internal procedures, staffing, collected data, a slow-start approach and the need to follow contact accuracy. Reasons were to be communicated. Separate additional information illustrated a possible phased introduction: perhaps one validation during the first year, potentially increasing towards quarterly checks, while the first pass could be spread across 12 to 24 months in batches. That example was explanatory rather than an operative obligation, but it showed a legitimate concern about implementation shock.

The strongest case for this system should be taken seriously. Search is costly. Failed delivery wastes time for a complainant and for an operator trying to determine whether a report ever arrived. Personal addresses are fragile. A role mailbox, presented consistently and readable by machines, can be maintained through staff changes. A phased first pass can prevent a sudden wave of checks from overwhelming members or AFRINIC. BTW’s analysis of abuse-contact economics helps put the benefit in the right terms: better contact data can reduce coordination costs without proving the allegation that travels through the contact.

NRS’s account of decentralisation and member accountability points in the same direction. A shared registry function can improve a common record without centralising the underlying dispute. Decentralised responsibility means the participant receiving the report remains responsible for its own procedure and applicable law, while competent public authorities remain responsible for matters within their lawful powers. The registry’s useful contribution is to ensure that the channel exists and remains reachable.

Heng Lu’s private-bookkeeper framework supplies the institutional boundary. AFRINIC has no legislative power to create general public duties, no administrative or regulatory power of a state agency, no judicial power to decide a contested case, no police or prosecutorial power to investigate and charge, no fiscal or taxing power, and no military or sovereign coercive power. A private service dependency does not acquire those powers because its records are essential. The absence of public authority does not make coordination unimportant; it makes precision about the coordinator’s task essential.

The outside-scope sentence in Draft 7 should therefore govern the ambiguous clauses. “Measures” and “response” can be read only as contact facts: did the designated channel accept a deterministic message, and did the holder complete the neutral confirmation needed to keep the record current? They should not become tests of whether the holder agreed with the complainant, investigated enough, used the preferred remedy or reached the outcome AFRINIC considered correct. A complaint about silence may justify checking whether delivery still works. It cannot convert the registry into an appeal body for the substance of the exchange.

A proportionate alternative is straightforward. Require one role address for each directly issued resource. Check it through a deterministic delivery challenge when the record is created, after a material update, following a confirmed bounce or on a predictable annual schedule. Send dated notice to the role address and administrative contacts. Give at least 30 days to correct the record. Issue a receipt for the correction and test again. Publish only a neutral state: reachable, temporarily failed, correction pending or unverified.

Most importantly, stop there. Do not test whether the allegation is true, whether the answer is good enough or whether remedial action is sufficient. Do not allow contact failure alone to affect recognised control, transfer access, voting, RPKI, reverse DNS, routing continuity or unrelated support. Alleged unlawful conduct belongs with competent authorities acting under applicable law. AFRINIC can provide accurate coordinates. It should not decide what happened at the destination.