Data Management and Sharing Plan
| Data types | |
| Metadata and standards | |
| Storage and preservation | |
| Sharing | |
| Access, security and ethics | |
| Roles |
This plan is written in the form funders ask for and is kept current as the programme moves from modelled to measured data. It governs every artifact the programme produces.
Data types
| Type | Description | Format |
|---|---|---|
| Synthetic corpora | Seeded forward-model spectra with generating parameters, planted corruptions, and split assignments | Columnar arrays and JSON metadata; generator source |
| Measured spectra | Raw and calibrated Raman acquisitions with instrument, settings, sample and site identifiers | Instrument-native export plus a columnar copy; JSON metadata |
| Derived quantities | Fitted band parameters with covariances, intensity ratios, stage assignments, refusal flags | JSON per record; CSV summaries |
| Models and evaluations | Trained inversion models, hyperparameters, held-out scores, probe results | Serialised model files; JSON evaluation records |
| Provenance records | Append-only ledger entries and instrument-journal receipts with Ed25519 signatures | JSON lines; chain heads and public keys published |
| Documents | Findings, preprints, plans | Markdown with SHA-256 digests; PDF |
Metadata and standards
Every record is described by the receipt schema of §2.2: actor, sequence number, timestamp, content digest, previous chain head, and signature. Spectra carry excitation wavelength, calibration readings before and after the session, and acquisition settings, following the conventions of Notations. File-level metadata uses plain JSON with documented keys; no proprietary container is required to read any artifact.
Storage and preservation
Code and pointers are kept in version control and binaries in a content-addressed store, so that one commit identifier reproduces a corpus and every model trained on it. Off-site copies are written to immutable, object-locked storage under an append-only credential, with daily, weekly and monthly snapshots and periodic read-back verification; encrypted offline copies are kept with the key stored apart from the media. A quarterly restore drill on a clean machine — clone, check out, pull, verify the ledger with zero failures, retrain and reproduce the reference score — is itself a release gate (Finding 2026.4). Retention is indefinite for released artifacts and for the ledger; working data is retained for at least five years after the end of the programme.
Sharing
Findings are shared on release at this site with their Markdown sources and digests, and are compiled into preprints deposited in a public repository with a digital object identifier. Synthetic corpora and their generators are held during the measured phase and shared on its completion, or earlier on request; their digests are published in advance so that a later release can be checked against the ledger. Measured data are shared subject to the agreements under which they were acquired and to export-control review, with a stated embargo where an agreement requires one. Provenance records — chain heads, public keys, and the verification procedure — are public.
Intended licences at release are Apache-2.0 for code, as with Matrix Scroll, and Creative Commons Attribution 4.0 for data and documents. Anyone using a released artifact is asked to cite the finding or preprint by number and version (Citing).
Access, security and ethics
The programme handles no personal data and involves no human or animal subjects. Data that a partner agreement or export-control review restricts are held on access-controlled systems and are not placed in public repositories. Release of any artifact is preceded by a check that no restricted material is included. Corrections to released data follow the Errata discipline: the artifact is re-released with an incremented version and a new digest, and the previous version remains available.
Roles
The author is responsible for this plan, for the ledger and its keys, and for release decisions. Where a partner laboratory acquires data, responsibility for the acquisition record rests with the acquiring laboratory and is signed as such; comparison of the two records is the programme's, and is published.