Foundations & architecture

    The layer a generator cannot emit — why the model is shaped as it is. Four foundational frames govern every term in the ontology: the UFO meta-category each class commits to, the three-graph separation that keeps meaning, validation, and advice apart, the six bounded contexts the standard partitions the industry into, and the inference / validation boundary that fixes what is entailed versus what is merely checked. Counts below are read at build time from the committed TTL corpus at source/03-standards/ontology/.

    UFO meta-category legend

    Every OPDA class commits to a Unified Foundational Ontology (UFO) stereotype with OntoClean meta-properties (rigidity, identity, dependence). This frame leads the reference — before any term table — because the stereotype determines how a thing may be classified, whether it may bear an identity criterion, and whether it may ever appear in a subclass tree. The crucial contract: a Role is anti-rigid and is never a subclass of a Kind (it borrows identity from its bearer); a Kind is rigid and carries its own identity criterion.

    UFO stereotypeOntoClean meta-propertiesOPDA classes
    Substance Kind rigid; carries an identity criterion; individuates its instances opda:Property, opda:LegalEstate, opda:RegisteredTitle, opda:Person, opda:Organisation, opda:Address
    Role anti-rigid; sortal — borne by a single Kind; never subclasses that Kind opda:Proprietor (borne by Person, founded by a Proprietorship Relator)
    RoleMixin anti-rigid; cross-sortal — borne by more than one Kind opda:Seller, opda:Buyer (Person OR Organisation), opda:Evidence (+ opda:evidenceType facet)
    Relator relational endurant; founded by an event; mediates two or more bearers; founds the roles it connects opda:Transaction (founds Seller / Buyer), opda:Proprietorship (binds Proprietor Roles to a RegisteredTitle)
    Mode / Quality existentially dependent on its bearer UPRN as a contingent Quality of a opda:Property (never the identity criterion)
    Quale / value structure a value in a quality region; a magnitude-plus-unit composite with no independent identity opda:MonetaryAmount (amount + currency), opda:RoomDimension (length + width); coded SKOS values (Quale-in-Region) drive the rest
    Information Object an information artefact existentially dependent on a generating activity — not a Substance Kind opda:EPCCertificate, opda:Survey, opda:Valuation, opda:Search, opda:Claim

    Source: opda-classes.ttl (the UFO meta-classes opda:Role, opda:RoleMixin, opda:Relator) and each module's rdfs:comment / skos:scopeNote, which record the Guizzardi 2005 and DOLCE (Masolo et al. 2003) provenance per class.

    Three-graph separation (ODR-0004 §3a)

    The ontology keeps three artefacts strictly orthogonal — OWL classes ⊥ SHACL shapes ⊥ annotations — and the separation is CI-enforced. This is the single most load-bearing architectural decision (ODR-0004, Council session-004): it prevents open-world class semantics from being contaminated by closed-world cardinality, and keeps advisory hints out of the normative model.

    GraphHoldsMust NOT contain
    Class graph
    opda-classes.ttl + module TBox
    owl:Class, rdfs:subClassOf, object/datatype properties, rdfs:domain/range (informative), labels & definitions no sh: triples
    Shapes graph
    opda-*-shapes.ttl
    SHACL node/property shapes that sh:targetClass the classes; cardinality, datatype, sh:in value constraints never owl:imports the class graph; no advisory annotations
    Annotation graph
    opda-*-annotations.ttl
    DPV personal-data co-annotations, AI hints (opda:aiHint), keyed to class/shape IRIs no sh: triples

    Shapes target classes via sh:targetClass; they never owl:imports them. No property carries an OWL cardinality restriction and a SHACL count constraint as if the two were equivalent. Inference materialises into a fourth, derived graph (…/pdtf/graph/inferred/entailment) and never pollutes the canonical three. The CI gate asserts, among others, that an ASK over the annotation graph for any sh:-prefixed predicate returns false — no SHACL triples may leak into the annotation graph.

    Six bounded contexts

    The standard partitions the UK property industry into 6 primary bounded contexts (DDD), each a skos:Concept in opda:BoundedContextScheme and each owning a SHACL overlay profile (ODR-0020). PDTF is the Published Language between them. Domain-term membership in a context is a derived, on-demand query, never materialised — and upstream conformist authorities (HMLR, MHCLG, DSIT) own no overlay, so they are linked by opda:consumesFrom and are not members of the scheme.

    ContextDefinition (with overlays)Steward
    Conveyancing Legal transfer of estate; overlays ta6/ta7/ta10/lpe1. Law Society; SRA; CLC; Society of Licensed Conveyancers; CILEx; CILEx Regulation.
    Estate Agency Marketing and sale of residential property; overlays baspi5/nts2. Propertymark; property portals (Rightmove, Zoopla); regulated by NTSELAT (transitioning to MHCLG).
    Mortgage Lending Lending secured against property; overlays fme1. UK Finance; BSA; regulated by the FCA.
    Property Data Services Aggregation and supply of property data; overlays con29R/con29DW/llc1/oc1/rds/sr24. COPSO.
    Property Technology Orchestration and system-to-system messaging spanning the industry contexts. No single regulator — orchestration layer across the industry contexts.
    Surveying Physical inspection and valuation; overlays piq. RICS.

    Read live from opda-contexts.ttl (6 concepts in the scheme).

    Inference / validation boundary (ODR-0029)

    The governing principle (ODR-0029, Council session-039): infer the relations whose closure the author wants materialised (subclass / subproperty hierarchy, inverse, symmetric, transitive) — and validate the constraints the author wants checked (domain, range, cardinality, datatype, identity keys) as SHACL. domain / range sit on the validation side; RDFS only ever put them on the wrong side.

    The entailment regime is understood as minimal / dormant: only rdfs:subClassOf type-propagation currently fires (0 triples on the schema, ~30 trivially recoverable on instance data); the other six rules are dormant carriers that activate when a real construct lands. The closure adds 0 triples over the flat schema — it is kept not for current output but as a governed boundary and negative consistency gate (no owl:sameAs, no spurious EPCCertificate ⊑ Property). The rule logic is frozen; the ruleset is a sound but RL-incomplete RDFS-Plus fragment (config/opda-rdfs-plus.rules), explicitly not an OWL 2 RL reasoner.

    The 7 frozen rules — the "Safe Group" (ODR-0025 §R1)

    #ConstructEntailment
    1rdfs:subClassOftype propagation up the class hierarchy
    2rdfs:subClassOftransitive subclass closure
    3rdfs:subPropertyOfproperty-value propagation to the superproperty
    4rdfs:subPropertyOftransitive subproperty closure
    5owl:inverseOfboth directions — (x p y) ⇒ (y q x)
    6owl:TransitiveProperty(x p y),(y p z) ⇒ (x p z)
    7owl:SymmetricProperty(x p y) ⇒ (y p x)

    Deliberately excluded (validated as SHACL, never inferred)

    Excluded constructWhy
    owl:sameAs materialisationirreversible identity propagation — catastrophic in master data (ODR-0005 §R5)
    owl:FunctionalProperty / owl:InverseFunctionalPropertytrigger owl:sameAs identity-merge
    owl:equivalentClass / owl:equivalentPropertyuncontrolled bidirectional propagation + transitive explosion
    rdfs:domain / rdfs:range type entailmenta property with N declared domains mis-types each subject as all N classes at once — validated as SHACL instead, never inferred

    This is what makes the regime RL-safe rather than RL or DL: no identity merging, no equivalence, no domain/range classification, no restriction reasoning. The corresponding domain / range constraints are validated by the ~273 sh:targetSubjectsOf → sh:class SHACL shapes (ODR-0029 R3) — see SHACL shapes.

    The implicit-Property crux and the classification doctrine are on Identity & classification. Browse the terms these frames govern: Classes, Properties, Vocabularies (SKOS). The full reference index is the Ontology section overview.

    Comments

    Loading comments…