# Reusing the world

*Building a meta-model · lesson 5 of 6 · ~15 min*

## What you will learn

How to assemble models from existing standards instead of writing everything: imports, the composition rules, and the 1180-standard external registry.

## Assembly beats authorship

The largest model built on this standard (a whole polity: 38 namespaces) reused or bound existing material for roughly two-thirds of its content. That ratio is the point: your domain is less unique than it feels. Addresses, countries, currencies, organizations, persons, invoices, skills, licenses: the world has already standardized these, often several times. Your model's value is in *your* semantics; everything commodity should be imported.

## The import machinery

External standards enter a model as **Semantic Packages**: versioned, isolated under `imports/`, with provenance preserved (which standard, which version, from where). Alignments between imported concepts and your local ones live under `mappings/` as **Semantic Mappings**: explicit, versioned, confidence-scored. Imported standards are a fixed special case of external mastership: the standards body is the master, releases flow in, you never edit your copy of schema.org.

## The composition question

The recurring design decision is deceptively small: when your `Employee` needs an address, is `Address`...

1. **a field** (`employee.address_line: string`),
2. **a nested model** (an embedded `Address` object),
3. **a reference** (a link to a sovereign `Address` object elsewhere)?

The standard answers with tests (identity: does it have a life of its own? authority: who governs its shape? reuse: do others need it?) and six composition mechanisms: EMBED, REFERENCE, MIX-IN, EXTEND, MAP, ANNOTATE. It also blesses when *duplication is correct*: snapshots (the address at shipping time SHOULD be frozen, not referenced), atomic value objects, and sovereignty boundaries. Knowing when copying is right is as important as knowing it is usually wrong.

## The external models registry

To save you the survey, the ecosystem maintains a catalogued registry of **1180 external standards** across 37 domains: semantic web, healthcare, finance, geospatial, legal, telecom, science... Each entry is classified by **compositional role** (is this standard a value-object? an entity model? a document format? infrastructure?) and a **default link type** (COMPOSE, REFERENCE, EXTEND, MIX-IN, ALIGN, EMBED). The workflow: before modelling any concept, search the registry; if a standard covers it, bind it with the suggested mechanism; model only the delta.

A **Connector Catalogue** goes one step further for the 45 most common joints (Address, Money, Quantity, Country, LEI, GTIN...): named, ready-to-use composition connectors between your model and the usual suspects.

## A worked instinct

The polity model binds ISO 3166-2 for territories, Akoma Ntoso and LegalRuleML for law, W3C DID and schema.org for persons, ESCO for skills, ISO 20022 for treasury, OWL-Time for the simulation clock: a dozen standards bodies in one Dimension, coherent because every binding declares its mechanism. None of those wheels got reinvented; all of that time went into the model's actual novelty (its value substrate).

## Key takeaways

- Model your delta; import the commodity. Two-thirds reuse is achievable at civilization scale.
- Field vs nested vs reference is a decided question with tests, six mechanisms, and lawful duplication.
- The registry (1180 standards, classified) plus the connector catalogue make "search before you model" a five-minute habit.

## Go deeper

- [Extension model: importing external standards](/spec/#02-architecture/Extension-Model.md)
- [Meta-model composition (ARCH-016)](/spec/#02-architecture/Meta-Model-Composition.md)
- [External models registry](/spec/#06-ecosystem/External-Models-Registry.md) · [Connector catalogue](/spec/#06-ecosystem/Connector-Catalogue.md)

Next: [Validation and evolution](06-validation.md)