Skip to content

Roadmap

openbim-dt now provides an ISO 23387 edition 2 contract layer, lossless semantic XML codec, typed views, built-in diagnostics, and CLI. This roadmap distinguishes that executable scope from schema-complete conformance and governance work.

Current baseline

Implemented in the 0.2 source line:

  • edition 2 and draft-placeholder namespace identities;
  • reusable GUID, multilingual text, reference, rational, data-type, unit/base, and concept contracts;
  • bounded XML parsing and semantic writing with unknown-content retention;
  • global DT family views and owned typed wrappers;
  • structured built-in validation diagnostics;
  • CLI inspection, validation, and rewrite operations;
  • synthetic fixture provenance, mutation probes, packaging leakage checks, API documentation, and OpenBIM facade integration.

Not implemented:

  • full XSD validation or clause-by-clause ISO 23387 conformance;
  • complete format-neutral owned representations for every optional ConceptType field;
  • byte-identical XML output;
  • ISO 23386 governance workflows;
  • ISO 12006-3 mapping;
  • a bSDD adapter.

Delivery principles

  1. Domain and wire layers stay separate. XML ordering is retained by the wire tree while reusable standard contracts remain application-facing values.
  2. Retain before interpreting. Unknown attributes, elements, order, and namespace evidence survive even when no typed accessor exists.
  3. Parsing is not validation. Structural decoding and semantic diagnostics are separate APIs and capability claims.
  4. Explicit editions and drafts. Namespace evidence is retained and draft inputs are never silently normalized into edition 2.
  5. Evidence-backed coverage. Claims require original or redistributable fixtures, positive/negative tests, and mutation probes.
  6. Dependency direction stays downward. Consumers may depend on DT; DT never depends on LOIN.

Milestones

1. Contract and codec baseline — implemented

  • Standard lexical values and reusable concept core.
  • Namespace-aware bounded parser and semantic writer.
  • Typed views/wrappers for all global edition 2 element families.
  • Unknown-content, comments, PI, CDATA, and order retention.

2. Built-in diagnostics — implemented subset

Current diagnostics cover namespace/root identity, required/duplicate/invalid GUIDs, ConceptType creation date/name requirements, reference identity, multilingual language tags, and data-type names.

Future work should add only diagnostics justified by public evidence and should keep warning/error policy stable once released.

3. Schema-complete model and validation — future

  • Measure XSD declaration/field coverage with a machine-readable inventory.
  • Complete owned contracts for optional lifecycle and provenance fields.
  • Evaluate an XSD validator without vendoring restricted schema material.
  • Publish exact edition and clause coverage; never equate parsing with conformance.

4. ISO 23386 and ISO 12006-3 boundaries — future

  • Model governance and dictionary mapping as explicit companion contracts.
  • Keep lifecycle policy and mapping provenance out of the wire codec.

5. Consumers — active

  • Use released DT contracts in LOIN only at ISO 7817-3 fields whose XSD types are imported from ISO 23387.
  • Evaluate dictionary/bSDD adapters independently.
  • Maintain acyclic dependencies and registry-versioned release order.

Standards material boundary

Purchased or otherwise restricted standards files remain local under references/. Documentation and release pipelines never copy that directory into Git, crates, fixtures, build artifacts, or GitHub Pages.