Skip to content

CONCEPT Cited by 1 source

RDF (Resource Description Framework)

Definition

RDF — Resource Description Framework — is the W3C standard graph data model for representing information as triples: (subject, predicate, object). Subjects and predicates are URIs (or blank nodes for subjects); objects can be URIs, blank nodes, or typed literals. A set of triples is an RDF graph; a set of named graphs is an RDF dataset (concepts/named-graph).

Everything else in the semantic-web stack — RDFS (class hierarchies), OWL (class expressions / reasoning), SHACL (shape constraints, concepts/shacl), SPARQL (query), and upper ontologies — is built on top of RDF.

Role on this wiki

The wiki's canonical RDF instance is Netflix UDA (systems/netflix-uda). Netflix chose RDF + SHACL as the foundation of UDA's knowledge graph:

"We chose RDF and SHACL as the foundation for UDA's knowledge graph." — Netflix UDA post (sources/2025-06-14-netflix-model-once-represent-everywhere-uda)

"Upper domain models are data. They are expressed as conceptual RDF and organized into named graphs, making them introspectable, queryable, and versionable within the UDA knowledge graph."

Netflix's use of RDF is notable because UDA is not a classical semantic-web deployment — it's an enterprise data- integration platform where RDF is the substrate that makes domain models, schemas, mappings, and system definitions uniformly manipulable as graph data.

Why RDF, not a property graph

The UDA post doesn't walk through this explicitly, but the implicit choice is:

  • Node / edge uniformity — in RDF, edges are themselves first-class resources with URIs, which matters when mappings and schemas need to be introspected + queried + versioned like data.
  • Standardised semantics — RDFS / OWL give subclass / subproperty / equivalence reasoning; Netflix reuses this for conservative extension.
  • Graph-dataset partitioning — named graphs are the load-bearing modularity primitive UDA needs (concepts/named-graph).
  • W3C + tooling — Jena (the Java RDF library), SPARQL, SHACL are mature and cross-implementation.

Gaps UDA had to fill

RDF alone wasn't enough; Netflix enumerates what had to be added on top (see systems/netflix-uda for full context):

  • Info model over named graphs — standard RDF doesn't prescribe how to organise named graphs, manage ownership, or define governance boundaries.
  • Enterprise-friendly validation — SHACL assumes globally unique URIs + one data graph; enterprise data has local schemas and typed keys (GraphQL / Avro / SQL).
  • Collaborative authoring — ontology tooling historically lacked the modular-contribution / ownership / safe-federation story that GraphQL Federation has.

Upper (systems/netflix-upper) is the layer that closes those gaps while staying on top of RDF.

Seen in

Last updated · 319 distilled / 1,201 read