linked.zone

Notes on RDF, ontologies and architecture models. Every note is a page and a resource.

Latest note · Argument · 27 Aug 2026

Divergence is a fact, not an error

Why a reviewed skos:exactMatch beats a reflexive owl:sameAs when two tool exports describe the same service, and how to keep the conflict as a first-class resource.

lzn:divergence-is-a-fact a lz:Note ; dct:subject skos:mappingRelation , prov:Entity ; dct:format text/html, text/turtle, application/ld+json .

All notes

6 notes

  • Argument

    Divergence is a fact, not an error

    Why a reviewed skos:exactMatch beats a reflexive owl:sameAs when two tool exports describe the same service, and how to keep the conflict as a first-class resource.

    dct:subject skos:mappingRelation , prov:Entity

  • Pattern

    Federating tool exports

    A seven-stage pipeline that keeps the three modes apart, with the failure each stage is there to catch.

    dct:subject lz:Topic

  • Pattern

    Shapes that fail on purpose

    Running drift shapes in CI against yesterday's graph, and reading the report as a diff over reality rather than a pass/fail.

    dct:subject sh:NodeShape

  • Argument

    Diagrams for presentation, not storage

    Why the drawing tool is the wrong system of record, and what a viewpoint actually is under ISO/IEC/IEEE 42010.

    dct:subject archi:Metamodel

  • Teardown

    Zachman cells are classification slots

    Reading the framework as a taxonomy, what that fixes in an OWL/SKOS encoding, and where the two standards genuinely do meet.

    dct:subject archi:Metamodel

  • Argument

    Five things OWL cannot say for you

    Cross-framework correspondence, non-transitive hierarchies, version-boundary statements, label semantics and controlled value sets, each with the shape of the statement that needs SKOS.

    dct:subject archi:Metamodel , skos:mappingRelation

In the queue

6 drafts

  • Baseline, target, and the graphs in between — Architecture states belong in separate named graphs as deltas, with the decision hanging off the graph rather than the elements.
  • Closed enumerations, open practice — Normative value sets and practice vocabularies need different mechanisms; sh:in for the first, SKOS for the second, never both for one slot.
  • Confidence is not trust — How sure you are that a statement is true and whether you are allowed to act on it are two axes, and one status field cannot carry both.
  • EA on a Page, checked against its sources — The widely circulated CSVLOD summaries misclassify artefacts the primary sources classify differently.