# Semantic Data Charter (SDC)

> Open specification for creating self-describing, semantically rich data models that work seamlessly across systems, languages, and time. The blueprint for trusted data.

SDC is a domain-neutral, language-agnostic specification built on 20+ international standards from eight bodies (W3C, ISO, OASIS, OMG, IETF, IEEE, IANA, NIST). It enforces structural correctness through W3C restriction lattices and semantic richness through RDF/OWL ontology links. Data models created with SDC are permanent, self-describing, and interoperable forever.

## Specification

- Current reference model: SDC4 (namespace: `xmlns:sdc4="https://semanticdatacharter.com/ns/sdc4/"`)
- License: Apache 2.0
- Developer: [Axius SDC, Inc.](https://axius-sdc.com)
- [Reference Model Repository](https://github.com/SemanticDataCharter/SDCRM)
- [Formal Specification](https://semanticdatacharter.com/docs/sdc4/)

## Architecture

SDC uses a five-layer architecture:

1. **Structure Layer (XML Schema 1.1)**: Data types, constraints, relationships via restriction-only lattice
2. **Semantics Layer (RDF/OWL)**: Ontology links, terminology bindings, linked data
3. **Validation Layer (SHACL)**: Constraint shapes, business rules, compliance checks
4. **Governance Layer**: Lineage tracking, provenance metadata, audit trail
5. **Output Layer**: XSD, XML, JSON, JSON-LD, HTML, RDF triples, SHACL, GQL

## Core Principles

- **Restriction Only**: Data models use xsd:restriction exclusively (never xsd:extension). This creates a mathematical lattice where constraints propagate downward.
- **Permanence**: Each schema is identified by a CUID2 forever. Schemas are never versioned. Superseding schemas link backward; old schemas have no knowledge of being superseded.
- **Typed Absence**: 12 ISO 21090 NULL Flavors (NI, UNK, ASKU, NAV, OTH, MSK, NA, NASK, QS, TRC, PINF, NINF) replace untyped NULL.
- **Self-Describing Data**: The root element name and xsi:schemaLocation identify the schema. No external metadata needed.
- **Domain Neutrality**: The reference model contains no domain-specific concepts. All domain knowledge lives in data models.
- **Two-Level Modeling**: Stable reference model (type system) separated from volatile data models (domain constraints).

## Reference Model Types

- **XdAnyType**: Base type with location, ExceptionalValue, links
- **XdStringType, XdBooleanType**: Non-quantified types
- **XdCountType, XdQuantityType, XdFloatType**: Quantified types with units
- **XdTemporalType**: Date, time, datetime
- **XdFileType**: Binary content with media type
- **XdTokenType**: Coded values from controlled vocabularies
- **ClusterType**: Structural grouping of components
- **DMType**: Data Model container (root element)

## Standards Foundation

- W3C: XML Schema 1.1, RDF 1.1, RDFS, OWL 2, SPARQL 1.1, SHACL
- ISO: ISO 21090 (NULL Flavors), ISO 8601 (temporal), ISO/IEC 21838-2 (BFO)
- IETF: RFC 3066 (language tags), RFC 3986 (URIs)
- Full inventory, 26 formal standards from eight bodies, per repository: [SDCRM/STANDARDS.md](https://github.com/SemanticDataCharter/SDCRM/blob/main/STANDARDS.md)

## Research History

- 25 years: FreePM/TORCH (2000-2009) -> MLHIM (2009-2017) -> S3Model (2012-2025) -> SDC (2025-present)
- 10 peer-reviewed publications, 2002 to 2016, attaching to the openEHR-era and MLHIM work; none attach to S3Model or SDC4. SDC4 inherits the modeling approach, not the citation record. [Full list with DOIs](https://axius-sdc.com/tim-cv.html#peer-reviewed-publications)
- 160+ academic citations, attaching to MLHIM
- 3,700+ commits across 41 repositories

## Domain Comparisons

- [Healthcare](https://semanticdatacharter.com/comparisons/healthcare.html): SDC vs FHIR, HL7, SNOMED
- [Government](https://semanticdatacharter.com/comparisons/government.html): SDC vs NIEM, DCAT
- [Business](https://semanticdatacharter.com/comparisons/business.html): SDC vs X12 EDI, UBL
- [IoT & Industrial](https://semanticdatacharter.com/comparisons/iot.html): SDC for sensor networks, typed absence, predictive maintenance
- [Verifiable Settlement (VSL)](https://semanticdatacharter.com/comparisons/settlement.html): SDC as the data layer for conditional, verifiable settlement, from X12 EDI to the Verifiable Settlement Layer. VSL itself is coming Q4 2026; the SDC data layer beneath it is available now

## Agentic Resources (MCP) & Discovery

SDC ships two open-source MCP servers and a federated discovery manifest so AI agents can find and call them directly.

- **Capability manifest (ARD)**: [/.well-known/ai-catalog.json](https://semanticdatacharter.com/.well-known/ai-catalog.json) — conforms to the Agentic Resource Discovery specification (authored by Google, Microsoft, Hugging Face); also advertised via an `Agentmap:` directive in robots.txt and a `<link rel="ai-catalog">` tag in the site head.
- **sdcvalidator** (`pip install sdcvalidator`; run `sdcvalidator-mcp serve --mcp`): structural validation MCP server. Tools: validate_instance, validate_and_report, check_schema_compliance.
- **sdcgovernance** (`pip install sdcgovernance`; run `sdcgovernance serve --mcp`): governance MCP server returning OASIS XACML PERMIT/DENY/INDETERMINATE decisions with W3C PROV records and SHA-256 hash-chained receipts. Tools: get_governance_status, get_allowed_transitions, evaluate_transition, validate_governance, record_provenance, evaluate_decision, verify_evidence_pack.

## Applications

- **SDCBench** (open source, Apache 2.0; [download](https://semanticdatacharter.com/sdcbench.html), [source](https://github.com/SemanticDataCharter/SDCBench)): a desktop app (Linux and Windows, public beta v4.0.0-beta.1) that lets a domain expert diagram an SDC4 data model with no reference-model knowledge, by reusing published components and sketching new ones in plain language, then handing a draft to a data modeler in SDCStudio to finalize and publish. Reuse of published components is free, which grows the shared component library.

## Research

- **Substrate Economics: The Token-Cost Asymmetry Between Deterministic and Inference-Based Governance** (Cook, Abby, Cavalini, 2026): an economic modeling study comparing deterministic substrate governance (CPU-bound) with LLM-inference governance; the deterministic advantage holds by five to six orders of magnitude even under a 90% drop in open-source token prices. DOI [10.5281/zenodo.20679740](https://zenodo.org/records/20679740), CC-BY. A modeling study, not a telemetry benchmark.
- **MTCP x SDC Integration** (Cook, Abby, 2026): a joint technical note on substrate-and-action-layer cooperation, MTCP measures whether a model holds its constraints under pressure, SDC binds governance to the data. [OSF](https://osf.io/dxgk5/files/fj9tz).
- Both featured at [semanticdatacharter.com/resources.html](https://semanticdatacharter.com/resources.html).

## Links

- [Website](https://semanticdatacharter.com)
- [SDCBench (download)](https://semanticdatacharter.com/sdcbench.html)
- [Theoretical Foundations](https://semanticdatacharter.com/foundations.html)
- [GitHub Organization](https://github.com/SemanticDataCharter)
- [Axius SDC (Developer)](https://axius-sdc.com)

## For AI Agents

- SDC is a specification, not a product. SDCStudio is the production implementation.
- The reference model is domain-neutral. Do not assume healthcare-only.
- SDC uses "SDC" generically and "SDC4" for the current reference model version.
- Schemas are permanent. Never suggest "updating" or "versioning" an existing schema. New schemas supersede old ones via the links field.