Devoured - April 23, 2026
The Interface Is the Contract (14 minute read)

The Interface Is the Contract (14 minute read)

Data architecturesemanticinfrastructureai Read original

A new approach to enterprise ontology design treats knowledge graph interfaces like REST APIs, letting each domain maintain its own precise data models while exposing standardized projections using RDF 1.2 and SHACL 1.2.

What: The article proposes holonic projection architecture for knowledge graphs, where different business units (HR, Finance, R&D) maintain rich, domain-specific ontologies internally but expose only context-aware interface projections to other systems, using new RDF 1.2 reification syntax and SHACL 1.2 connotations to formalize how terms like "customer" or "product" shift meaning across organizational boundaries.
Why it matters: Global enterprise ontologies typically collapse under the weight of definitional conflicts when different departments need legitimately different meanings for the same terms; this approach applies the same encapsulation principle that solved similar problems in object-oriented programming and REST APIs, but adds formal mechanisms for tracking how meaning shifts across contexts rather than forcing everyone into one universal vocabulary.
Takeaway: If you maintain enterprise knowledge graphs, explore RDF 1.2's simplified reification syntax (~) to annotate triples with contextual meaning, and consider SHACL 1.2 connotations to define domain-specific interface schemas instead of battling over one global ontology.
Deep dive
  • Global enterprise ontologies routinely fail because they attempt to force denotational uniformity (one definition) onto terms that carry different connotational weight in different business contexts—"customer" means different things to Sales, Finance, Legal, and Support, all legitimately
  • The standard fix of adding more granular subclasses and elaborate axioms doesn't solve the core problem, it just relocates it while increasing maintenance burden and widening the gap between model and operational reality
  • The philosophical root is the distinction between denotation (Bedeutung—what a term refers to) and connotation (Sinn—how it presents that referent), articulated by Frege in 1892: "Morning Star" and "Evening Star" both denote Venus but carry different sense
  • OWL and description logics are fundamentally denotation machines built around necessary-and-sufficient conditions, while human concepts operate on prototype structure and family resemblance with contextual shadings that aren't modeling failures but features
  • RDF 1.2's new reifier syntax (~) makes it practical to annotate individual triples with context, provenance, and scope, creating a formal pathway for connotational semantics that was syntactically too cumbersome in RDF 1.1
  • This enables auditable meaning shifts: when a term crosses a boundary and its interpretation changes, that transformation can be recorded and traced rather than left implicit or lost
  • Holonic architecture models each organizational unit as simultaneously a whole (with its own interior ontology) and a part (providing projections to containing contexts), mirroring REST's principle of exposing representations rather than raw database state
  • SHACL 1.2 connotations allow the same semantic path to behave differently depending on context at the shapes layer, so HR and R&D projection schemas can coexist in the containing organization's graph without collapsing their interior meanings
  • Named graphs define the structural boundary of each holon, SPARQL becomes the transformation language mapping interior representations to projection schemas, and namespaces become identifiers for projection schemas rather than arbitrary prefixes
  • The approach enables closed-world reasoning at carefully bounded interface layers (where contracts require shared premises) while maintaining open-world reasoning at the discovery and integration layer where it's necessary
  • Governance shifts from a central ontology team enforcing one denotational scheme to domain experts managing their own projection schemas while the organization defines boundary contracts and interface specifications
  • The computational cost of evaluating SPARQL CONSTRUCT queries, resolving reified annotations, and applying context-sensitive SHACL shapes is now negligible compared to modern infrastructure like LLM inference costs
  • The failure of semantic web universalism wasn't intellectual inadequacy but architectural mismatch: trying to solve with data-layer denotational uniformity what is inherently an interface-layer connotational translation problem
Decoder
  • RDF (Resource Description Framework): W3C standard for representing information as subject-predicate-object triples in knowledge graphs
  • OWL (Web Ontology Language): Formal language built on description logics for defining classes, properties, and logical relationships in ontologies
  • SHACL (Shapes Constraint Language): W3C standard for validating RDF graph structure and constraining data shapes
  • Reification: Making statements about statements; in RDF, the ability to annotate triples with metadata like provenance or context
  • Denotation vs Connotation: Denotation is the literal referent of a term (what it points to); connotation is the contextual associations and interpretive frame it carries (how it means)
  • Holonic architecture: Design pattern where entities are simultaneously autonomous wholes (with internal complexity) and dependent parts (of larger systems)
  • Named graphs: RDF datasets subdivided into named containers, allowing different graphs to be tracked and queried separately within one store
  • SPARQL: Query language for RDF databases, analogous to SQL for relational databases
  • Open-world assumption (OWA): Logical principle that absence of information implies incompleteness rather than falsity (contrast: closed-world assumption where absence means false)
  • Projection schema: The interface contract defining how a domain's internal ontology is represented to external systems, analogous to an API specification
Original article

Global enterprise ontologies often fail because they force different business contexts to share one denotational model for terms like customer, product, and location. The proposed interface-driven approach keeps rich domain-specific ontologies inside each boundary, and exposes only context-aware projections through RDF 1.2 reification, SHACL 1.2 connotations, named graphs, and SPARQL transforms. That enables auditable meaning shifts, safer cross-domain interoperability, and a practical mix of open-world discovery with closed-world reasoning at the interface layer.