Skip to content

Landscape: Sourced Findings

Type: Non-normative

Status: v0.2 (2026-09-28)

This file keeps sourced findings separate from SMF proposals. Each entry states what the work is, what it covers that matters to SMF, and what it leaves out. Citations are in bibliography.md.

Operating and management frameworks

ITIL 4

  • Service management framework (PeopleCert/AXELOS).
  • Relevant to SMF: an interoperability target, not a model to copy. Semantic Assurance failures and definition changes can flow into existing ITSM incident and change processes.
  • Gap: the managed object is IT services. Business meaning is not a managed object.

DAMA-DMBOK

  • Data management body of knowledge (DAMA International). Version 2 is current; the v3.0 project kicked off in June 2025 and aims to incorporate AI, cloud and modern data platforms. [DAMA 3.0]
  • Relevant: data governance, metadata management, reference and master data, stewardship roles.
  • Gap: organized as knowledge areas; not designed as independently adoptable practices. No explicit resolution or interpretation assurance. Commercially published.

EDM Council DCAM

  • Capability assessment model for data management programs.
  • Relevant: governance and authority structures, capability scoring.
  • Gap: assessment of data management capability, not an operating model for meaning.

ISO/IEC 42001 and NIST AI RMF

  • AI management system standard and AI risk framework.
  • Relevant: accountability, testing/evaluation, lifecycle governance of AI systems. Semantic Assurance can supply semantic test evidence to these.
  • Gap: neither treats business meaning as a governed object.

Semantic standards and methods

ISO/IEC 11179 (metadata registries)

  • Registration-based model for metadata items with registration authorities, stewardship and submission roles.
  • Part 6 defines registration statuses. Lifecycle statuses: Incomplete, Candidate, Recorded, Qualified, Standard, Preferred Standard, Retired, Superseded. Documentation statuses: Historical, Application. [NSW MDR]
  • Relevant: authority, identity and lifecycle status model SMF should adopt rather than reinvent.
  • Gap: no contextual resolution or consumer assurance.

OMG SBVR and the semantic/speech community distinction

  • SBVR (OMG) separates meanings (concepts) from their expressions (designations), enabling multilingual and multi-vocabulary models. [SBVR]
  • The SBVR-aligned lexicon distinguishes a semantic community (shares understanding of concepts, not necessarily the same terms) from a speech community (also shares the same terms). [BRC Lexicon]
  • Relevant: direct prior art for "concept vs. term" and scoped vocabulary.

W3C SKOS

  • Vocabulary for concept schemes: preferred/alternative labels and mapping relations (exactMatch, closeMatch, broadMatch, narrowMatch, relatedMatch).
  • Relevant: SMF mapping types should align with SKOS where possible.

Domain-Driven Design

  • Bounded contexts make meaning explicitly local to a model and team. Context maps describe relationships between contexts (e.g., shared kernel, anti-corruption layer, published language).
  • Relevant: strongest prior art for "authoritative does not mean globally singular."

Business Semantics Management

  • Research-derived method (VUB STARLab; De Leenheer, Meersman) with two cycles: semantic reconciliation among communities, and semantic application committing information systems to agreed patterns. [BSM]
  • Relevant: closest historical precursor to SMF's discover, define, authority and bind flow. Associated with the origins of Collibra.

FAIR-IMPACT semantic artefact governance

  • Governance and management framework for semantic artefacts in research: lifecycle phases (conceptualization, encoding, publication, maintenance), 18 governance aspects, organizational role domains, versioning and deprecation. [FAIR-IMPACT]
  • Relevant: a working example of an operating model for semantic artefacts, but for research communities, not enterprises.

EKGF Maturity Model

  • Open maturity model (CC BY-SA 4.0) from the Enterprise Knowledge Graph Foundation, an OMG-managed community. Four pillars (business, organization, data, technology), five maturity levels, and explicit positioning as grounding for generative AI. [EKGF]
  • Relevant: closest open peer. Useful structural reference for an SMF maturity model.
  • Gap: assumes knowledge graphs as the vehicle; maturity scoring rather than adoptable practices.

Representation and technology

Apache Ossie (formerly Open Semantic Interchange)

  • Apache incubating spec for exchanging semantic models across analytics, AI and BI platforms. [Ossie]
  • Core building blocks: datasets, fields, relationships, metrics; ai_context hints; custom extensions. [Datapace]
  • Omits trust metadata: confidence, lineage, freshness, provenance, verification/approval records. [Datapace]
  • Industry commentary argues portability and governance are distinct problems: the format describes a metric but does not govern or execute it. [Strategy]
  • Spec 0.2.0.dev0 (draft): one model per document, no bundle or cross-model references; a metric is name, expression (SQL per dialect), description, datatype, ai_context (instructions, synonyms, examples) and custom_extensions (vendor name plus a JSON string). [Ossie spec]
  • No field for owner, approval, status, effective date, perspective, comparability or data-contract links; two meanings of one metric can only be two metric names. [Ossie spec]
  • Relevant: preferred interchange target; SMF governance lives above it. Field-level mapping and a linking convention are in the reference.
  • Gap: ai_context hints are per model, so cross-model agreement (which perspective, when to ask) has nowhere to live inside the format; custom_extensions is the only slot for a pointer out.

Open Data Contract Standard (ODCS)

  • Bitol (Linux Foundation) specification for data contracts; v3 documented. [ODCS]
  • Relevant: reference for dataset-level contracts and implementation bindings.

dbt Semantic Layer / MetricFlow and similar

  • Metric definition and query generation technology with entities, measures, metrics and dimensions.
  • Relevant: common implementation target for SMF measurement contracts.

Vendor agent context layers

  • Example: Snowflake describes an "agent context layer" combining semantic models, ontology/identity layers, operational playbooks, provenance and decision memory, with explicit attention to disambiguation and confident wrong answers. [Snowflake]
  • Relevant: confirms market demand for resolution and provenance for agents.
  • Gap: platform architecture, not a vendor-neutral operating model; limited treatment of roles and testing.

W3C PROV

  • Provenance data model. Relevant: substrate for claim assurance records.

LF AI & Data Foundation projects (reviewed 2026-09-28)

Source: LF AI & Data projects.

Egeria (incubation)

  • Open metadata and governance type system, APIs and interchange protocols (Apache-2.0).
  • Glossary terms with synonyms, preferred and replacement terms, IsA, and UsedInContext, which "distinguish[es] between terms that have the same name but different meanings depending on the context." Semantic assignment links terms to implementations. [Egeria]
  • Relevant: closest prior art for SMF's Modeling practice and for context-dependent meaning. A natural store for SMF documents.
  • Gap: a metadata model, not resolution behavior or consumer testing.

Bitol: ODCS and ODPS (incubation)

  • ODCS v3.2.0 and ODPS v1.1.0 (released 2026-09-08) add a context block with instructions, verified statements and constraints for AI agents; semanticType (column, measure, dimension); synonyms; enums; a deprecated flag. [ODCS 3.2]
  • Relevant: strong overlap with SMF's context package at the dataset level. SMF should reference ODCS rather than duplicate it.
  • Gap: dataset-scoped; no cross-dataset concepts, classification rules, scoped authority, conflicts or resolution states.

OpenLineage and Marquez (incubation)

  • Open lineage standard and API; custom facets (with _producer and _schemaURL) can attach to runs, jobs and datasets. [OpenLineage facets]
  • Relevant: provenance substrate for ClaimTrace; candidate home for an SMF semantic facet.

Monocle (sandbox)

  • Tracing framework for GenAI applications. [Monocle]
  • Relevant: source of agent traces to check against SMF TestCase expectations.

Unity Catalog OSS (sandbox)

  • Open catalog for data and AI. Metric views are Databricks-platform features, not part of the open-source project; no semantic layer shipped Ossie import/export as of August 2026. [Atlan UC]
  • Relevant: possible host for SMF documents and discovery.

Amundsen (graduated)

  • Data discovery and metadata engine. Relevant: Discovery practice harvesting source.

OpenSharing (Linux Foundation, launched 2026-06-10)

  • Open protocol for sharing data and AI assets, including agent skills, contributed by Databricks. [OpenSharing]
  • Relevant: possible cross-organization carrier for semantic context (Interchange).

Agent frameworks: BeeAI (incubation), RYOMA (sandbox), Open Platform for Enterprise AI (sandbox)

  • Relevant: realistic SMF consumers to test resolution and clarification behavior. [RYOMA]

DataPractices (graduated)

  • Manifesto of values and principles for data teamwork. Relevant: precedent for a principles-first open framework hosted by LF AI & Data.

Naming collisions checked

  • "SemOps" / "Semantic Operations Framework" is used by SemOps.ai (an AI technology offering) and by Semantic Arts.
  • "Open Semantic Contracts" was considered and set aside (v0.3): it described one component, not the discipline.
  • "Semantic Management Framework" is the chosen name. A formal name and trademark check is still to be done.
  • Related but distinct: Apache Ossie (formerly Open Semantic Interchange), ODCS, and the semantic-mesh GitHub project (domain contracts), which has not yet been reviewed in detail.
  • "MetricProof" is an existing MIT-licensed package (AtomicGlance).

Market signals (non-authoritative)

  • "Semantic drift" and "context drift" are now common vendor terms for AI failure modes (e.g., AtScale, Atlan). [AtScale] [Atlan]
  • These describe the symptom Semantic Assurance addresses, without defining an operating model for it.