Namespaces and identifiers

First distinguish three states

ODR-0065 selects the namespace policy explained here; implementation in ontology artefacts remains pending. Selecting that policy does not rename the existing ontology, regenerate the Property Pack candidate or make these example addresses live. Read an identifier together with its corpus and decision status.

A term called Report in all three spaces is not thereby one term. Replacing a base string would mint a different IRI without establishing correspondence, preserving an identity criterion or approving the new definition. Any eventual transition needs its own scoped decisions and evidence.

Choose the family by meaning and authority

The selected topology adapts semantic-modelling’s scheme to opda.org.uk. Its root kernel and common boundary are deliberately different. The kernel describes how the ontology is classified or governed; common contains domain meaning genuinely shared across contexts.

There is no additional /core/ layer: the cross-cutting kernel is /ns/ itself. Nor does a directory called schema determine a semantic namespace. For a root-level term, justify all three commitments: cross-cutting scope, metamodel meaning and no property-domain meaning. A reusable domain class belongs in its justified context or common boundary, not in the kernel because it seems foundational.

Upstream facet families include subject, dataclass, lifecycle, gov, vol, reg and valuechain. Their names are reserved against context-token collisions; reservation does not introduce seven new OPDA modelling obligations. Only selected facets and justified values are admitted. Subject classification answers “what is this about?”; context ownership answers “who governs this meaning?”

Use the established contexts, with stable tokens

The tokens reuse abbreviations already recorded in the Property Pack candidate manifest. Their selected bases are new policy; the candidate’s existing full IRIs remain unchanged. These are flat peers, not nested branches of a larger business-domain tree.

The first six rows are the established OPDA bounded contexts. Common is a separately governed shared boundary; DBT Smart Data is scheme scope, not a seventh OPDA bounded context. Namespace allocation does not create relationships between these boundaries or settle which candidate terms each should own.

Treat sv as a stable token after adoption, not a phrase to regenerate from a display label. Renaming a working group, translating its label or changing a prefix alias does not rename /ns/sv/. Likewise, do not insert a source-system name beneath it merely because two applications use different vocabulary within the same governed context.

Prefixes abbreviate; governance explains

With sv: explicitly bound to https://opda.org.uk/ns/sv/, sv:Inspection abbreviates the first example. Another document could bind survey: to that same base and name the same term. Conversely, the candidate can use sv: with a different base and name a different term. The expanded IRI matters, not the letters before the colon.

This is especially important for opda:. In this policy it is a convenient alias for https://opda.org.uk/ns/; existing schema-derived material uses it for https://opda.org.uk/pdtf/. Never silently reinterpret that older binding. Show full IRIs or distinct aliases when discussing both corpora together.

VANN’s preferredNamespacePrefix and preferredNamespaceUri record preferred namespace bindings explicitly on the admitted vocabulary’s logical owl:Ontology resource. They do not prove that an organisation controls a domain, approve a definition or establish that two contexts share meaning. The context-map profile separately identifies a context, an identifier system it governs and, for a namespace-based system, the explicit namespace value. A context may govern multiple identifier systems.

That distinction makes governance queryable without treating a familiar prefix or substring as evidence. A namespace declares the intended naming boundary; the authoritative registry and reviewed model establish governance. An identifier-system resource is not interchangeable with the context resource or a named graph.

Stable identifiers are not file or release addresses

Keep four things distinct: a term IRI names a vocabulary resource; an individual IRI names a particular entity; an ontology IRI identifies a logical ontology; a document address identifies a document or representation. A named graph has its own graph name. None is automatically identical to a repository filename, directory or page route.

One namespace may span several files; one file may contain terms from several namespaces. Splitting a Turtle file, moving a source directory or redesigning this site therefore does not rename its terms. A graph can package statements from several contexts without becoming their semantic owner. The namespace policy is not a rule to create one graph or one document per context.

Nor does every release require a new term namespace. A corrected label can describe the same term; a changed definition may alter meaning and require a separate reviewed decision. Preserve established references, document deprecation and replacement where justified, and never quietly reuse an old IRI for a different meaning. Instance continuity likewise needs domain reasoning: correcting a report record is not automatically a new inspection.

The version inside the current candidate base belongs to that candidate’s identity. It does not prescribe /ns/v1/ for the selected policy. owl:versionInfo records version information; owl:versionIRI, when used, identifies a particular ontology version without renaming every term. Release identifiers, immutable version documents and delivery locations require their own explicit contracts; this policy does not invent their paths. Upstream’s dated operational graphs and versioned KB documents are scoped exceptions, not templates for OPDA domain identifiers.