In a dataspace, policies define who may use which data, for what purpose, and under which obligations. Policies are part of contract offers and agreements, not optional extras. A policy agent (policy decision point) evaluates requests against rules and returns allow/deny before data leaves the provider.

Eclipse Dataspace Components use a policy engine on the control plane to attach constraints to assets and contract definitions. During negotiation, both sides must agree on compatible policies before contract finalization and data transfer. ODRL-based expressions are common for machine-readable rights: ODRL model (W3C).

For API-level runtime authorization (for example SPARQL endpoints), MAT-X can use Open Policy Agent (OPA) with Rego policies. OPA evaluates request attributes (method, path, bearer token, etc.) and keeps policy logic separate from application logic.

Further reading Open Policy Agent (Wikipedia) - policy-as-code overview EDC policy engine - connector policy evaluation ODRL Information Model (W3C) - rights and obligations vocabulary IDS RAM - data sovereignty - policy enforcement in IDS architecture Gaia-X trust framework - compliance, credentials, policy

A semantic data model describes what data means, not just how it is stored. Resources are linked to shared vocabulary concepts so machines and humans can interpret and query data consistently across systems.

On the web of data, meaning is represented with graph standards:

RDF expresses facts as subject-predicate-object triples. OWL 2 defines classes, properties, and logical constraints. SHACL validates whether instance data matches expected structures.

An ontology formally specifies a domain: classes, object properties, datatype properties, and subclass axioms. Reference ontologies like PMD Core ontology anchor project-specific extensions so new terms stay interoperable. MAT-X pipelines (Autoencoder, Agent-TOM) extract or learn candidate classes but ground exports under reference rdfs:subClassOf parents.

Further reading Semantic Web (Wikipedia) - standards and linked data Ontology - information science (Wikipedia) - ontology role in knowledge representation RDF 1.1 Concepts (W3C) - graphs, IRIs, literals, datasets OWL 2 Web Ontology Language Primer (W3C) - classes, properties, reasoning Apache Jena documentation - RDF storage, SPARQL, Fuseki Gaia-X interoperability docs - semantic interoperability in federated ecosystems

A data space (dataspace) is a trusted ecosystem for sharing data across organizational boundaries while each participant keeps sovereignty over assets. Participants publish discoverable offerings, negotiate contracts, and exchange data under agreed rules instead of copying everything to a central warehouse.

The International Data Spaces Association (IDSA) defines reference architecture and certification for trustworthy data spaces. Gaia-X promotes a federated, open European cloud/data infrastructure with common trust, identity, and interoperability building blocks. Initiatives like Catena-X show sector-scale dataspace operation in practice.

Technically, many sovereign dataspaces use a connector pattern: each participant runs software that exposes a catalog, negotiates usage contracts, and transfers data only after policy checks pass. Eclipse Dataspace Components (EDC) implements this connector stack as open source. MAT-X local participant setup is based on EDC plus a semantic RDF layer.

Further reading Data space (Wikipedia) - concept overview EDC documentation - architecture and APIs Gaia-X technical documentation - trust framework and interoperability IDS Reference Architecture Model (RAM) - layers, roles, and building blocks