Problem thread

Graphs, tables and property graphs

What is the right shape for connected data: tables, triples, or graphs whose nodes and edges carry properties?

issue
data model
status
open
steps
11, 1968–2026
threads
all problem threads

Knowledge representation drew knowledge as networks and frames, while databases settled on tables and entity–relationship diagrams. The Web adopted RDF's triples, and industry adopted labelled property graphs, whose edges can carry their own properties. In 2019 the communities met at a W3C workshop; since then ISO has published GQL for property graphs, and RDF 1.2 adds statements about statements, but the two graph models remain separate.

Lineage

  1. 1968
    precedent writing
    Semantic Memory

    Stores word meanings as a network of linked concepts: knowledge drawn as a graph. [src]

  2. Jun 1970
    A Relational Model of Data for Large Shared Data Banks

    Organize data as relations, that is tables, independently of how it is stored. [src]

  3. Jun 1974
    A Framework for Representing Knowledge

    Frames: structures of slots with default values for stereotyped situations. [src]

  4. Mar 1976
    The Entity-Relationship Model: Toward a Unified View of Data

    Model data as entities and the relationships between them, as a unified view over the database models of the day. [src]

  5. 6 Jun 1997
    Meta Content Framework Using XML

    Models information as a directed labelled graph: nodes, and arcs that are triples of source, target and label, also called properties. [src]

  6. 22 Feb 1999
    settles standard ★ seminal
    RDF

    The W3C's model for data on the Web: statements of subject, predicate and object, each identified by a URI. [src]

  7. 2000
    proposes standard
    Topic Maps

    ISO's alternative, built around topics, associations and occurrences, with expressive power similar to RDF's. [src]

  8. 2007
    proposes project
    Neo4j

    A database of labelled property graphs, where nodes and relationships carry labels and key–value properties, queried with Cypher. [src1][src2]

  9. 2019
    poses writing
    Report of the W3C Workshop on Web Standardization for Graph Data

    Brings the SQL, property-graph and RDF communities together. They agree that mappings between property graphs and RDF should come first, back a property-graph query language alongside SQL, and split 16 to 12 on extending RDF. [src]

  10. 12 Apr 2024
    settles standard
    GQL

    ISO publishes GQL in April 2024, a standard query language for property graphs, partly derived from Cypher. [src]

  11. 7 Apr 2026
    responds standard
    RDF 1.2

    Adds triple terms, so a statement can be the object of another statement: the extension the 2019 workshop discussed for statements about statements. [src1][src2]

Open now

Under each question, the answers people have put forward or built. The catalogue gives no odds on them.

  • Can RDF data and property-graph data be mapped to each other without losing meaning? [src]
    The 2019 workshop agreed that a standard should cover these mappings before any interoperability between query languages.
    • One underlying model serves both, and users pick the query language.
      Amazon Neptune's team proposes OneGraph, a single graph model that embraces RDF and property graphs.[src]
    • RDF grows to say what property graphs say. RDF 1.2
      RDF 1.2's triple terms let a statement be described by further statements, much as a property graph puts properties on an edge.[src]
  • Will tools and publishers adopt RDF 1.2's triple terms? [src]
    The core RDF 1.2 specification reached W3C Candidate Recommendation in April 2026.

Decision points

Dated moments when a decision or a deadline can change the outcome.

  1. 30 Apr 2027
    RDF & SPARQL Working Group charter ends charter

    The W3C group finishing RDF 1.2 and SPARQL 1.2 is chartered until this date.[src]

How to contribute