# TLA-VIS — plant visualisation grammar **Status:** 1.0.0 — ratified, all 330 terms published **Licence:** Vocabulary + specification **CC-BY-4.0** · reference implementation **Apache-2.0** · "TLA-VIS" name and conformance mark retained **Date:** 2026-09-07 **Generated by:** `npm run foundation:tla-vis` → `public/foundation-wing/tla-vis-vocabulary.json` **Related:** [`PLANT_CONDITION_GRAMMAR.md`](./PLANT_CONDITION_GRAMMAR.md) · [`GEOMETRY_IMPROVEMENT.md`](./GEOMETRY_IMPROVEMENT.md) --- ## Is this a real thing? Yes, and the gap is genuine. Nobody owns this space. | Existing standard | Covers | Licence | Why it does not solve this | |---|---|---|---| | **Plant Ontology (PO)** | Plant anatomy, morphology, development stages | CC-BY-4.0 | Built for genomics and phenotyping. Describes a *leaf primordium* for gene annotation, not how to render a crown. | | **Darwin Core (TDWG)** | Occurrence, specimen, taxon records | CC-BY | Where a plant was seen, not what it looks like. | | **TRY / AusTraits** | Measured functional traits | CC-BY variants | Numbers per taxon. No form vocabulary, no individual state. | | **glTF / OpenUSD** | Geometry and scene transport | Royalty-free / permissive | Taxon-agnostic. They carry a mesh; they do not know what a phyllode is. | | **MaterialX / OpenPBR** | Material and shading description | Apache-2.0 | Surface response, not botanical form. | | **IFC / buildingSMART** | Built asset data | Open | Has a generic `IfcGeographicElement` for planting. Effectively a dot on a plan. | The missing layer is a **controlled vocabulary that maps botanical form and individual condition to renderable parameters**. Plant Ontology stops where rendering starts; glTF starts where botany stops. Nothing joins them. We have already built that layer by accident. It is scattered across `factory-species-identity.mjs`, `genusBloomPriors.js`, `barkPriors.js`, `leafTraitModel.js` and the engine enums. TLA-VIS is the act of naming it. --- ## What is in it Generated from the live engine, so the document cannot drift from the renderer. | Module | Terms | What it fixes | |---|---:|---| | `habit` | 33 | Growth form, crown shape, trunk habit, branching gradient, phyllotaxis | | `foliage` | 150 | Leaf templates, twig clusters, posture, render mode, frond type | | `lamina` | 77 | Lamina class, margin, apex, base, venation — the bridge to flora prose | | `surface` | 14 | Bark profile, leaf material class | | `reproductive` | 16 | Bloom type, inflorescence arrangement | | `condition` | 26 | Water, decline, wound taxonomy, response — **new, and nobody else has it** | | `temporal` | 8 | Season, age stage | | `provenance` | 6 | Basis, trait provenance — required for conformance | | **Total** | **330** | plus 16 measured (numeric) slots | The `condition` module is the differentiator. Every other module is a tidier version of vocabulary that exists somewhere in the industry. Condition — a named, gated, occlusion-aware grammar for what has *happened* to an individual plant — does not exist in any published standard we can find. --- ## What it is not Scope discipline is what makes a standard adoptable. - **Not a file format.** Geometry ships as glTF or USD. TLA-VIS rides in `extras` / custom schemas. - **Not a shading model.** Surfaces resolve through standard PBR. - **Not a taxonomy.** Names come from the relevant flora authority. - **Not a genomics ontology.** Plant Ontology already covers anatomy for phenotyping; we should map to it, not compete. - **Not a replacement for TLA-185.** Different instrument, different question. --- ## Relationship to TLA-185 They are siblings, not versions. | | TLA-185 | TLA-VIS | |---|---|---| | Question | What is *recorded* about this taxon | How does this plant *read* | | Consumer | Revit, procurement, climate screening, counsel | Renderers, game engines, Studio+, visualisation | | Unit | Field | Term | | Numbering | By field count (169 → 185) | Semantic version | | Individual state | None | Core | **Why not number it by term count.** 169 and 185 are records standards where the field set is deliberately bounded. A vocabulary grows by design — adding a wound kind must not force a rename. TLA-VIS uses semantic versioning and reports its term count as a property. This is a decision to confirm, not a fait accompli. --- ## Licensing — the part that decides whether this works **TLA-185 is currently CC-BY-NC-ND-4.0. Copying that onto TLA-VIS would kill it.** This is the single most consequential decision in the document, so the reasoning is spelled out. ### Why NC-ND fails for a grammar - **ND (NoDerivatives) forbids extension.** Standards spread through profiles — a NZ flora profile, a vendor profile, an arboricultural profile. ND makes every one of those a licence breach. You would be the only permitted author of every extension, forever. - **NC (NonCommercial) makes implementation legally unsafe.** No commercial vendor will ship a feature whose specification forbids commercial use. NC does not stop competitors; it stops *implementers*, which is the opposite of what a standard needs. - **The precedent runs the other way.** Plant Ontology started on CC-BY-ND-3.0 and **moved to CC-BY-4.0**. Darwin Core is CC-BY. glTF, USD and MaterialX are all permissive. There is no widely adopted NC-ND technical standard, because there cannot be one. ### What copyright actually protects here A list of terms is thin copyright — you cannot copyright a system, a method, or facts. What you *can* hold: 1. The **specification text** — its expression, structure, worked examples. 2. The **curated definitions** — the botanical judgement in each entry. 3. The **reference implementation** — real code, ordinary software licence. 4. The **trademark** — and this is the strongest lever by far. ### Adopted posture Ratified 2026-09-07. This is the decision, not a proposal. | Asset | Licence | Reason | |---|---|---| | Vocabulary JSON | **CC-BY-4.0** | Must be implementable. Attribution keeps the name attached. | | Specification text | **CC-BY-4.0** | Extensions and translations are how it spreads. | | Reference implementation | **Apache-2.0** | Includes a patent grant, which vendors' lawyers look for. | | "TLA-VIS" name and conformance mark | **Trademark, retained** | Anyone may implement; only conformant implementations may claim the badge. | | Curated species records, meshes, stills | **Unchanged, commercial** | The moat. | **The moat is the data, not the grammar.** An open grammar makes the curated archive *more* valuable, because every implementer becomes a potential consumer of records that speak it. A closed grammar protects a vocabulary nobody uses. --- ## Conformance Three levels, so implementers can start cheap. | Level | Requirement | |---|---| | **1 — Consumes** | Recognises terms; renders something plausible; degrades unknown input to the default rather than inventing. | | **2 — Round-trips** | Reads and re-emits without term loss. | | **3 — Honest** | Enforces expression gates (refuses terms a growth form or taxon cannot express) and carries `basis` on every condition claim. | Level 3 is the interesting one and the reason to publish at all. It is a machine-checkable definition of *not lying about a plant*, which is the thing the industry currently has no way to state. --- ## What publishing actually requires Being honest about the work, because "publish a standard" hides a lot. - [x] Vocabulary extracted from the engine and generated reproducibly - [x] Condition module designed with expression gates - [x] Machine-readable artefact (`tla-vis-vocabulary.json`) - [ ] **Stable identifiers** — every term needs a permanent URI (`TLAVIS:0000123`, resolvable). Renaming a term after publication is a breaking change forever. - [ ] **Definitions.** 330 terms currently have names, not prose definitions with citations. This is the bulk of the remaining work and it is botanical, not technical. - [ ] **Deprecation policy.** How a term dies without breaking consumers. - [ ] **RFC process.** Sibling to the existing `federation/rfc/` path. - [ ] **Conformance test suite.** A level-3 badge is worthless without one. - [ ] **Governance.** Who arbitrates. A standard controlled by one commercial party is adopted reluctantly; a Foundation-stewarded one with an external advisory seat is adopted properly. - [ ] **Mapping tables** to Plant Ontology and AusTraits. Cheap credibility, and it prevents the "you reinvented PO" objection. - [ ] Counsel review of the licence split. Realistically: the vocabulary is a week of engineering and **several months of botanical definition writing**. The definitions are the standard. The term list is just an index. --- ## Decisions taken | Decision | Outcome | Date | |---|---|---| | Licence split | Open — CC-BY-4.0 vocabulary and spec, Apache-2.0 reference implementation, trademark retained | 2026-09-07 | | Scope of first publication | **All 330 terms** as TLA-VIS 1.0, not condition alone | 2026-09-07 | | Naming | `TLA-VIS` with semantic versioning; term count is a reported property, not the name | 2026-09-07 | ## Still open 1. **Governance** — Foundation-stewarded with an external advisory seat, or wholly internal for now? A standard controlled by one commercial party is adopted reluctantly. 2. **PO mapping** — before or after the definitions land? Cheap credibility, and it heads off the "you reinvented Plant Ontology" objection. 3. **Stable identifiers** — the term list is published but terms do not yet have permanent resolvable URIs. Renaming after publication is a breaking change forever, so this should land soon. 4. **Definitions.** Publishing all 330 terms means 330 names are now public with only module-level prose behind them. The definitions are the standard; the term list is an index. This is the bulk of the remaining work and it is botanical, not technical.