Registry federation research note
Question
How can agent registries exchange or project state without converting federation into unbounded trust inheritance?
Working position
Registry federation should preserve provenance, competence, scope and conflict information. A receiving registry may make another registry’s records visible, but visibility must not imply local authority unless the receiving governance domain explicitly accepts that authority for the relying context.
Failure modes
| Failure mode | Risk |
|---|---|
| Authority laundering | A weak or compromised registry becomes trusted because a stronger registry republishes its records |
| Provenance loss | Relying parties cannot determine where a claim originated |
| Conflict flattening | Simultaneous but inconsistent records are hidden behind a single projected status |
| Scope expansion | A record valid in one context is treated as valid in another |
| Revocation delay | Receiving registries continue serving stale authority after source revocation |
| Privacy amplification | Federation creates wider correlation than the original registry intended |
Minimum federation metadata
A federated projection should preserve at least:
- source registry identifier;
- source record identifier;
- selected source checkpoint or version;
- source competence claim;
- receiving registry interpretation;
- time requested and time evaluated;
- known conflicts;
- later material events when reconstructing historically;
- whether the receiving registry accepts, rejects, warns on or merely republishes the source claim.
Experiment candidate
Use two registries and one relying party:
- Registry A publishes an active agent record.
- Registry B consumes or projects the record.
- Registry A suspends the agent.
- Registry B is evaluated before and after receiving the suspension event.
- The relying party asks whether the agent is authoritative at requested time T1 and T2.
The experiment should show whether Registry B can disclose provenance, freshness, conflict status and its own acceptance boundary.
Expected output
- a sequence diagram;
- example federated records;
- positive and negative vectors;
- recommended conformance assertions;
- documentation update for the interoperability programme.
Open issues
- Whether federation requires a common transport profile;
- how to express source competence without requiring a single global trust framework;
- whether imported records should use separate identifiers or canonical source identifiers;
- how much retained evidence is sufficient for later dispute;
- whether privacy controls should be source-declared, receiver-enforced or both.