Discussion input

SDC and the Context Graph four facets

A bind-at-source mapping, written for the W3C Context Graphs Community Group.

This maps the Semantic Data Charter onto the Context Graph Protocol's four-facet model and proposes SDC as a bind-at-source arm for the tax-preparation benchmark. It is a discussion input, not a proposal to change CGP. SDC is one input that arrives at a boundary with its Context Facets already resolved. Version 0.2, September 2026.

The one-line claim

CGP measures whether a boundary crossing's Context Facets are LIT or DARK. SDC is a way to make Meaning, Structure and World arrive LIT, because it binds them into the record at authoring time. In CGP's own terms: SDC lights the facets the gauge measures.

The mapping

CGP facet	Question it answers	What SDC binds into the instance	State at the crossing
`Data`	What crossed the boundary?	the value itself, as an SDC-typed element	present by observation
`Meaning`	What does the Data signify?	the concept and definition, plus the ontology reference, bound to the element	LIT
`Structure`	How is it encoded, validated, generated?	datatype, units, range, cardinality, rounding rule, value set, schema version, expressed in XSD 1.1 and emitted as SHACL	LIT
`World`	Which external referent, authority, time?	jurisdiction, effective policy version, temporal validity and source lineage in W3C PROV	LIT

SDC binds all three Context Facets at authoring time, so a gauge observing SDC-bound data records low Dark Context by construction. A bare ontology-linked record, by contrast, lights Meaning but leaves Structure and World DARK: the reference range, the rounding rule and the governing authority never crossed the boundary with the value. That DARK gap is the failure mode the tax-preparation benchmark is built to measure.

Worked example: Form 1040, the standard deduction

A pipeline hands a downstream agent the value 14600.

Without bound context

Data: 14600
Meaning: DARK. Deduction, credit or wages?
Structure: DARK. Dollars? Whole-dollar rule?
World: DARK. Which tax year, which filing status, whose authority?

One stray assumption, applying it to a different tax year or filing status, silently corrupts the return. This is the compounding effect of broken Context.

As an SDC component

Data: 14600
Meaning: LIT. "Standard deduction," with its definition and tax-vocabulary reference.
Structure: LIT. USD, whole-dollar, non-negative, taken only when not itemizing (a cross-line constraint), emitted as SHACL.
World: LIT. US federal (IRS), Tax Year 2024, filing status single, the revenue procedure that set the amount, with PROV lineage.

The agent can now compute, refuse or ASK with the evidence in hand. Nothing has to be reconstructed at the boundary.

How it plugs in: the SHACL adapter

The group's Semantic-Alignment repository states the goal directly: organizations with existing ontologies, knowledge graphs and constraint languages should be able to express their definitions as inputs to the Meaning facet without translation loss. SDC already emits SHACL, along with XSD 1.1, OWL, RDF, JSON-LD and W3C PROV, from a single bound component. So the path is direct:

SDC component -> SHACL + PROV -> facet-table adapter -> Data / Meaning / Structure / World, LIT at source

This is an open-source adapter contribution (Apache 2.0). It does not change CGP and does not introduce a required data model. SDC is one input among many, arriving without translation loss.

Which committee that work belongs to under charter v2 is a question rather than an assumption. Format-level concerns appear to sit with Serialization and Specification, but the group is best placed to say.

Proposed benchmark contribution

Add a bind-at-source arm to the tax-preparation benchmark:

Baseline A: raw values, no bound context (facets DARK).
Baseline B: ontology-linked values (Meaning LIT, Structure and World DARK).
Arm C: SDC-bound components (Meaning, Structure and World LIT).

Measure Dark Fraction and the declared quality, accuracy, cost and latency across arms, then stress each by removing or corrupting context: a wrong tax year, a changed rounding rule, a swapped filing status. The hypothesis is that Arm C starts at the lowest Dark Fraction, holds accuracy, and degrades least under the broken-context stress, because the interpretive prerequisites were bound at the source rather than reconstructed at the boundary. If it does not, that is a more interesting result than if it does.

What SDC is not claiming

SDC is not a context graph and not a replacement for CGP. CGP is the gauge at the boundary; SDC is one bound source that makes facets arrive LIT.
SDC does not decide interpretation for another participant. It binds the authoring party's own declared constraints, consistent with CGP's semantic sovereignty and gauge-not-judge principles.
The honest difference is timing. CGP can resolve a DARK facet at runtime, through the Governor's ASK and negotiate path. SDC resolves it at authoring time, so less remains to negotiate at the boundary. The two compose: SDC lowers Dark Fraction at the source, and CGP measures and handles whatever remains.

SDC (the Semantic Data Charter) is an open-source specification (Apache 2.0) for self-describing, structurally governed data, with meaning and its constraints bound to the record.