Getting started
Establish ownership, understand the standards boundaries and set up a reproducible local environment.
Repository review: . Public operational reference; acceptance remains to be recorded by the receiving owner.
First day: establish a safe baseline
- Identify the receiving owners. Agree who owns the programme, standards decisions, website, AWS operations, Microsoft workspaces, participant data, communications and NotebookLM. Record a deputy and escalation route privately. A repository maintainer does not automatically inherit each of these authorities.
- Receive the repository and source archive. Confirm read/write access to the correct GitHub repository and its Actions settings. Work in the existing
maincheckout. Rungit status --shortandgit log -5 --oneline; preserve other people's work. Record the starting commit and the currently deployed release identity. - Read the contributor contract. Begin with
AGENTS.md,README.md,DESIGN.md,Makefileand the relevant accepted ADRs. Some old guidance describes earlier hosting or time-limited executor arrangements; compare it with current code and record conflicts before relying on it. - Build from the locked dependencies. Follow the local setup instructions, then run
make testandmake build. Keep the results and inspectgit status --shortagain for generated changes. A site build requires neither AWS credentials nor a production database. - Inspect the public experience. Check the six main destinations, Property Pack pages, search and resource downloads. Use the approved Chrome connection and existing OPDA profile for local browser work. Reading is public; optional member sign-in serves participation features.
- Complete a supervised release and recovery exercise. Use a small authorised change, observe the CI gates and deployment, verify the published commit, then demonstrate the recovery procedure in a controlled setting. Agree an incident contact and confirm an operational alert actually reaches its recipient.
The transfer is ready for acceptance when a successor can repeat these steps with their own approved access and can explain the remaining open items. The final checklist is in the acceptance checklist.
Understand the four bodies of work
The site explains work with different origins and authority. Keep these distinctions in pages, model outputs, presentations and generated media.
| Body of work | Role | Authority boundary |
|---|---|---|
| PDTF schema | Existing Digital Property Pack JSON Schema, overlays, dictionary, glossary and supporting implementation material. | Attributed third-party technical input. It does not determine SPDTF business meaning or demonstrate working-group agreement. |
| Schema-derived ontology | The separate draft linked-data extraction from that schema corpus, with RDF/OWL, SKOS, SHACL, mappings and provenance. Stable technical identifiers retain /pdtf/**. | A technical derivation with its own evidence and validation. It is neither the PDTF schema itself nor the SPDTF standard. |
| SPDTF | The Smart Property Data Trust Framework being developed collaboratively across domain practitioners and stakeholders. | In development and non-normative. Human groups determine meaning through recorded governance; publication does not confer statutory or standards authority. |
| Property Pack ontology candidate | An independent, greenfield delivery intended as a future SPDTF component, covering a closed source scope of 451 required items. | Machine-proposed and non-normative. Source items are coverage requirements, not 451 predetermined RDF properties. Technical determination and subsequent domain review remain separate authority stages. |
Accepted ADR-0063 establishes domain-led bounded contexts; ADR-0066 and ADR-0067 establish the Property Pack scope and architecture; ADR-0075 establishes its accelerated component role; ADR-0077 establishes the PDTF input boundary. ADR-0065's AI-assisted workflow and ADR-0068's wider lifecycle and ratification policy remain proposed. Do not turn their suggested process into an adopted rule through paraphrase.
Use Programme for purpose, Governance for authority, Modelling for method, Development for drafts and inputs, Groups for participation and Resources for evidence. The separate association website and this knowledge base have different publishing responsibilities.
Local setup and prerequisites
Use Node.js 22.12 or newer within Node 22 to satisfy Astro's engine requirement and match CI, Corepack and the pnpm@11.17.0 version pinned in package.json. For ontology work, add Python 3.11 and Java 21. The pinned Jena/Fuseki 6.1.0 binaries require Java 21; older repository comments saying Java 17+ are stale. Docker is only needed for optional Skosmos. Marketing asset regeneration requires local ffmpeg.
# In the repository root, on main
corepack enable
make install
make dev
# Repeatable static-site checks
make test
make build make install uses the frozen pnpm lockfile. make dev selects a free loopback port from 4330–4339 and reuses an existing project server. Use the exact address it prints. Do not run competing dev processes against the shared Vite cache. make preview builds and serves the production output.
# Additional setup for opda-gen
python3.11 -m venv tools/opda-gen/.venv
make ontology-install
# Use the same Python for root-level catalogue/candidate commands
source tools/opda-gen/.venv/bin/activate
python3 --version
java -version
# Provision the pinned Jena distribution through its existing resolver
# Downloads and checksum-verifies it when no installation is found
python3 -c "from opda_gen.jena_shacl import resolve_jena_shacl; print(resolve_jena_shacl())" Keep this Python environment active for the model recipes. Root-level package scripts invoke python3, while make ontology-install and the opda-gen targets select the repository environment themselves. Confirm that the version checks show Python 3.11 and Java 21.
The Jena resolver checks OPDA_JENA_HOME, then PATH, then the local cache. Verify that any existing override is the intended 6.1.0 distribution. The Property Pack validator does not download Jena: it needs executable riot, arq and shacl together under JENA_HOME/bin, the repository's .jena/ cache or PATH. If the resolver uses an external installation, make that same installation discoverable to the candidate validator.
# Optional model-query development; mutates local dataset/projections
# Ctrl-C stops these owned services before a later make build-data
make serve-data make serve-data starts Fuseki on port 3031, clears and reloads its opda dataset, rewrites src/data/ontology-model.json and public/data/ontology-graph-elements.json, then starts the local API on port 3002. Inspect generated diffs afterwards. Confirm that no other Fuseki process owns the selected port or shared local data directory; stop the owned interactive services before running make build-data. Fuseki is separately downloaded and checksum-verified when required.
Some full ontology gates need the upstream schema archive at source/03-standards/schemas/; obtain the revision pinned by .github/workflows/site-release.yml rather than an arbitrary latest checkout. Catalogue regeneration also needs its original Property Pack workbook and matching schema evidence, as described in Models and validation.
The development server exposes selected local evidence files for the resource viewer. The production build does not copy the whole archive. A resource that works locally is therefore not proof that its approved production download exists.