Search RPD Archives
[rpd] Reminder: Deadline for Last Call – Today 23:59 UTC.
Tshepo Masuku
TshepoMasuku26 at hotmail.com
Fri Jul 31 09:40:14 UTC 2026
Dear PDWG,
I wish to raise a separate implementation concern arising from the Impact Assessment after reviewing the Soft Landing, Recovered Space and Priority Policy.
The assessment records no impact on WHOIS, RDAP, MyAFRINIC, NetSuite, NMRP, RPKI, or IT. However, it also requires AFRINIC to introduce an automated cleanup and monitoring tool and allows six months for implementation of that workflow.
These positions require clarification.
If the proposed tool merely observes recovered resources, then it is unclear how it will implement or enforce the movement of resources from quarantine into the available pool.
If the tool can change a prefix’s status, release it from quarantine, place it in reserve, or make it eligible for allocation, then it is a new operational control system. In that case, the conclusion that there is no systems or IT impact appears incomplete.
The more important concern is that the policy does not define the decision logic that the automation will execute. The community has not been told:
#which data sources determine whether a prefix is considered cleaned;
#whether transitions are automatic or require human approval;
#what happens when data are missing or inconsistent;
#whether the system fails closed when a check cannot be completed;
#who may override an automated result;
#how changes to the tool’s decision rules will be reviewed;
#or how every release decision will be audited and reproduced.
These are not minor implementation details if they determine whether scarce IPv4 space becomes available for allocation. They are part of the substantive rule.
Before the proposal advances, the Impact Assessment should be updated to identify the tool as an affected system and publish at least:
its functional scope and authoritative data sources;
the exact decisions it may make automatically;
its human approval and override controls;
its failure and rollback behaviour;
its versioning and change-control process;
test cases demonstrating the intended outcomes; and
an auditable record of each decision affecting resource availability.
No prefix should become allocable solely because an unpublished internal workflow classified it as clean.
Otherwise, the community approves broad policy language while the real allocation rules are written later in software outside the PDP.
That is not implementation of policy. It is policy being created through implementation.
For this reason, I do not believe the current Impact Assessment is sufficient, and I remain opposed to the proposal advancing as written.
Regards,
Tshepo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.afrinic.net/pipermail/rpd/attachments/20260731/16b45b17/attachment-0001.html>
More information about the RPD
mailing list