Roadmap¶
Axioval aims to make validation requirements portable without making packages executable or privileging one vendor, model format, or repository owner.
This roadmap describes intent, not a compatibility guarantee. Milestone contents may move as conformance tests expose missing foundations.
Now: trustworthy authoring substrate (0.1.x)¶
- Typed Pkl authoring modules and deterministic normalized JSON.
- Static-first manifest inspection and repository-confined evaluation.
- Fail-closed package, definition, parameter, selector, and snapshot binding.
- Reusable object/property vocabulary with external-schema names.
- Optional exact property-container qualification.
- Rich public documentation and a complete vocabulary-to-instance tutorial.
- Publish immutable
v0.1.0after external consumer review.
Next: portable capability governance (0.2.x)¶
- Define a capability registry with stable semantic contracts and conformance vectors, not engine implementations.
- Specify deterministic diagnostic/result interchange.
- Add unit/quantity system identifiers and conversion contracts.
- Define conflict, missing, unsupported, and invalid fact states explicitly.
- Add package dependency identities, checksums, and lock data.
- Publish migration tooling for normalized schema changes.
Then: ecosystem and registry (0.3.x)¶
- Launch
axioval/registrysubmission and discovery workflows. - Validate immutable repository revisions in isolated workers.
- Publish signed validation attestations and normalized artifact hashes.
- Index applications' declared capability support.
- Add package compatibility matrices without ownership-based trust.
- Support external vocabulary packages shared by many rulesets.
Stable contract (1.0.0)¶
A 1.0.0 release requires:
- documented schema-version negotiation and migration rules;
- multiple independent package authors and checking-engine consumers;
- a public conformance suite with fail-closed negative vectors;
- stable capability and diagnostic semantics;
- reproducible package validation; and
- an explicit deprecation and security-response policy.
Open design questions¶
- How should canonical concepts reference bSDD, IDS, and non-IFC vocabularies without making one service mandatory?
- Which duplicate-property cases are conflicts versus precedence policies?
- How should type-level versus occurrence-level property sources be represented?
- Which template-composition features lower cleanly to declarative normalized data without becoming a hidden programming language?
- How should localized diagnostics and remediation guidance be standardized?
Non-goals¶
- Shipping package-provided executable checking logic.
- Owning IFC parsing, geometry kernels, or application runtime IRs.
- Treating folders or property sets as universal semantic hierarchies.
- Giving Axioval-owned repositories special registry trust.
- Claiming a documentation example is a normative DIN or IFC compliance rule.