Skip to content

ODRL Policies

The W3C Open Digital Rights Language (ODRL 2.2) expresses permissions, prohibitions and duties in a form many policy systems read. In FAIR² it is an export format: a dataset's terms are stated in its governance blocks, and an ODRL policy is derived from them for systems that consume ODRL.

Changed in v1.7.0

Up to v1.6.3 this page described embedding a hand-written ODRL policy in fair2.json as the statement of a dataset's terms. Governance blocks replace it: they are what governance services enforce, and they carry terms ODRL cannot.

Which form is authoritative

  1. fair2:access and the other governance blocks are the canonical form. When they are present, a consumer reads them.
  2. An ODRL policy is read only when a document has no fair2:access — for example one produced by a system outside FAIR².

A governance service may publish the ODRL rendering of a dataset's governance (for instance from a policy endpoint), marked as derived from fair2:access.

Mapping

FAIR² governance ODRL
Access tier (L0–L2) odrl:Permission with a credential-tier constraint
+G community governance odrl:Duty to obtain community approval
+S environment odrl:Constraint on the attested environment
+P purpose binding odrl:purpose constraint; per-column purposes target the column, with DPV purpose IRIs
+J jurisdiction odrl:spatial constraint
aiTraining: "prohibited" odrl:Prohibition of odrl:derive
Embargo odrl:dateTime constraint
Data use agreement odrl:Duty to accept the agreement
Ethics training odrl:Duty to hold the training

The tier, approval, environment, agreement and training terms on the ODRL side belong to the FAIR² governance model's ODRL profile. They appear only in exported policies and are not part of the fair2: vocabulary this specification defines.

What does not survive the export

The mapping is lossy. ODRL has no equivalent for, among others:

  • the difference between a declared and a verified purpose;
  • the people exempt from an embargo (exemptSubjectIds);
  • accepted identity providers and identity-assurance levels;
  • the agent ceiling tier and agent policy;
  • temporal phases;
  • query governance: granularity, group sizes, linkage policy;
  • Indigenous governance bodies and cultural protocols;
  • the approvers and approval threshold of community governance.

A system that enforces a dataset's terms must therefore read the governance blocks, not the ODRL rendering.