Summary

  • At the RIPE 93 Database Working Group on 29 October 2026, Kaveh Ranjbar proposed optional referrals from RIPE NCC's RDAP service to a resource holder's server for more-specific registration records. He asked for an impact analysis and working-group discussion, not an adoption vote.
  • Leo Vegoda said he broadly supported the idea, then named two checks: the delegated service must keep working, and it must return the same answer to RIPE NCC and every other client. The second check addresses a possible split view; the meeting record documents no instance of one.

The handoff changes who answers

An RDAP query can follow an address to the most specific record RIPE NCC holds. Ranjbar's published talk abstract argues that a holder may keep more-specific records on its own server, beyond that endpoint. His proposal would let a holder put its RDAP base URL on the covering object, send subordinate lookups there, and add a visible link announcing the delegation. RIPE NCC would validate the service and withdraw the referral if it failed.

That sounds like a directory forwarding a caller to the right desk. The fine print is that the desk is now a source of registration answers. A redirect carries the registry's authority into a server operated by the holder. It can improve access to more-specific data without forcing those records into the central database. It also asks clients to trust what the holder-run service says after the redirect.

Vegoda, speaking from the floor, separated two failure modes. An unavailable server is easy to notice: a query fails. An inconsistent server can answer perfectly when RIPE NCC checks it and give a different answer to an ordinary client. Vegoda wanted a check that everyone receives the same answer. This was a design warning, not an accusation that a named operator had done it. Ranjbar agreed to discuss the details.

The difference matters because a green uptime check cannot establish truthful, uniform answers. A service could be available all day and still treat clients differently. The proposal's fourth part mentions ongoing validation and withdrawal on failure, but the session published no test plan for cross-client consistency, no withdrawal threshold, and no decision to deploy the feature. Those are the next records to look for. The handoff is easy to draw with an arrow; the harder diagram is who verifies the desk at the arrow's end.

Sources: RIPE 93 Database transcript; RIPE 93 talk abstract; Leo Vegoda's RIPE Labs profile.