Implementation Registers Architecture
Please note that this specification is suitable for pre-production pilot implementations.
The implementation governance layer is organised as four registers — the technical roots of the UNTP governance tree. Together they catalogue the software, extensions, conformity schemes, and identifier infrastructure that every UNTP implementation draws on.
Registration is a continuous process, not a one-shot certification event. An implementer publishes a DID, signs a registration Verifiable Credential, and exposes their service or credentials at a stable URL. From that point onward, the UNTP registrar agent (did:web:registrar.uncefact.org) crawls each registered implementation on every observation cycle, validates conformance against published tests, and writes signed observations back to the register. Status reflects observed reality, not declared intent — critical at the scale UNTP is designed for, where millions of independent implementers cannot be manually audited.
The four registers
| Register | Catalogues | What the implementer publishes | What the agent verifies |
|---|---|---|---|
| SWI — Software Implementers | Software vendors and the build versions they ship | DID + registration VC + an issuingSoftware block in every credential the software issues | Signature, schema conformance, vocabulary conformance, optional binding-credential alignment, sample evidence per software version |
| EXT — Community Extensions | UNTP credential extensions defined by member organisations | DID + registration VC + extension schema, JSON-LD context, and vocabulary at stable URLs | Schema/context/vocabulary content hashes, UNTP context import, no UNTP redefinitions, all terms resolved, samples validate |
| CVC — Conformity Vocabulary Catalogue | Conformity schemes (audit, assessment, certification programs) | DID + registration VC + a UNTP-conformant CVC vocabulary at a stable vocabularyURL | Vocabulary URL reachable, CVC schema conformance, profile/version URI immutability, topic URI resolution, standard/regulatory alignment reachability |
| IDR — Identifier Schemes | Identifier schemes for products, facilities, organisations, members, and key assets | DID + registration VC + a publicly reachable resolver service URL | Registry URL reachable, identifier pattern compiles, resolver template resolves, optional UNTP IDR conformance, optional DIA issuance verification |
Each register has its own governance document with the full publishing checklist, validation rules, and lifecycle. Per-register detail is linked in the sections below.
Why this model
Self-attestation does not scale to millions of implementers — there is no way to manually audit a million conformance claims, and even sampling becomes unreliable. The agent-driven model gives verifiers two things self-attestation cannot:
- Continuous integrity, because every observation is a freshly-signed Verifiable Credential, and any drift from the registered configuration (silent schema changes, broken resolvers, unsigned credentials, forged software claims) is detected on the next crawl cycle.
- Cryptographic auditability, because observations are append-only and signed by the registrar agent's DID — no party can quietly modify history, including UNECE.
The implementer's job stops at "publish your DID, issue a registration VC, expose your service at a stable URL." Everything downstream — observation, validation, drift detection, status updates — is the agent's job.