# MODAVIS identity and organ link coverage

This addition links the reference catalog to the public MODAVIS Pipe Organ
Dataset (POD) **1.6.0**, retrieved on **10 September 2026**. It does not change
historical acquisition evidence or establish new physical-organ identities.
The reference remains version 0.1.0, an updated unpublished draft under
reserved DOI 10.5281/zenodo.22693457.

## Results by scope

| Scope | Entries | Explicit public identities | Entries with accepted organ links | Distinct related organs |
| --- | ---: | ---: | ---: | ---: |
| Retained discovery catalog | 3,600 | 897 | 316 | 171 |
| Historical acquisition inventory | 119 | 119 | 22 | 21 |
| Broader collection-evidence listing | 123 | 122 | 23 | 22 |
| Ten analyzed VPOs | 10 | 10 | 0 | 0 |

These overlapping scopes must not be added together. An entry with multiple
accepted targets counts once in the linked-entry column; organ identifiers are
deduplicated within each scope. Different editions can refer to the same organ.
These are not counts of complete downloaded sample sets.

All 897 returned public records are marked independently distributed; 316
(35.23%) have accepted organ links, reproducing the supplied Navigator filter and
its 171 distinct organs. The other 2,703 discovery rows are **not returned in this
public projection**, not necessarily without identifiers elsewhere in MODAVIS.
Subordinate entries are not assigned their parents' links by this crosswalk.

The inventory has 22/119 linked entries (18.49%). The broader listing has 23/123
(18.70%); the additional linked entry is VPO-401, supported here only by a native
definition directory label. This is not proof of an additional independently
acquired sample set. UNASSIGNED-001 has no direct public catalog-ID mapping.
All ten analyzed VPOs have virtual-instrument identities but no accepted organ
links in this release; this does not invalidate their samples or spectral results.

## Exact join and acceptance rules

1. Join retained `entity_id` to API `canonical.externalIdentifier` by exact equality.
2. Count identities only for exactly one matching public record. Preserve missing
   and ambiguous cases. Do not fuzzily match names, towns, builders or producers.
3. Count organ relationships only when `targetOrganMdvsId` is non-null and
   `resolutionState` is `accepted_canonical_organ_relation`.
4. Retain all returned relations, including unresolved evidence-only rows, with
   relation type, source confidence, resolution state and source-row hash.
5. Do not inherit links through parent catalog identifiers. Any future inherited
   mapping requires a separately documented derivation and distinct counts.

The identity and relationship evidence partly derives from the same catalog.
This is a crosswalk to published identities, not independent validation of all
original catalog facts. Confidence values retain their source meaning and are
not treated as newly calibrated probabilities.

## Relationship meaning

Published relation types include sampling from a real organ, historical
reconstruction, inspiration and generic pipe-organ relationships. Preserve the
specific relation instead of calling every link a recording-source match.
Organ identity alone does not establish sampled configuration, recording date,
edition identity, download completeness, file integrity or redistribution rights.
Do not combine identity coverage and HTTP availability into a single claim of
whole-corpus reproducibility.

## Persistent references

Virtual-instrument w3id URLs are copied from the API's `identifierUri`.
Physical-organ URLs apply the same published entity policy to explicit target IDs:
`https://w3id.org/modavis/entity/{public-identifier}`. The policy was checked
against a linked organ; individual URI resolution was not exhaustively tested.

Example: **VPO-414**, Adlington Hall Free Demo:

- Virtual instrument: https://w3id.org/modavis/entity/A961-F8D6-6
- Physical organ: https://w3id.org/modavis/entity/2HXM-25TN-N
- Relation: `vmiorgreltype:sampled_from_real_organ`.
- Version-labelled export:
  https://data.modavis.org/dataset/pod/version/1.6.0/entity/A961-F8D6-6.modavis.jsonld

Persistent entity URLs may resolve to later descriptions. The version-labelled
exports and archived snapshot bind this crosswalk to POD 1.6.0. A MODAVIS package
or version identity is not proof of byte identity with an acquired archive.

## Frozen evidence and offline replay

`modavis-navigator-1.6.0-snapshot.json.gz` preserves exact UTF-8 response bodies for
nine public-list pages, the filtered count response and release-context responses
before and after retrieval. Each has its URL, timestamp and SHA-256. The compressed
snapshot is hashed in `modavis-crosswalk.json` and the package manifest.

Release fields are preserved distinctly: `releaseVersion=1.6.0`,
`dataProfile=pod-1.5-public`, `identityLedgerReleaseVersion=1.5.5`. They are not
interchangeable dataset version numbers. No source audio is fetched.

The list query is:

```text
https://navigator.modavis.org/api/virtual-instruments?limit=100&offset=0&independently_distributed=all
```

Subsequent offsets are 100 through 800. The public projection returned only 897
independent records despite requesting all distribution states. Coverage of an
unavailable internal projection is not inferred.

The filter count query is:

```text
https://navigator.modavis.org/api/virtual-instruments?limit=1&organ_link=linked&independently_distributed=yes
```

User-facing filter:
https://navigator.modavis.org/virtual-instruments?organ_link=linked&independently_distributed=yes

Replay offline in a working copy:

```sh
python3 modavis_crosswalk.py --package . \
  --snapshot modavis-navigator-1.6.0-snapshot.json.gz \
  --output modavis-crosswalk-replayed.json
```

The replay should match `modavis-crosswalk.json` byte-for-byte with the supplied
serialization. The package verifier also recomputes the mapping without creating
extra files. To capture a future release, work outside the archived package,
pass `--fetch --release VERSION` and choose a new snapshot path. Capture fails
on an unexpected release or inconsistent pagination/release context.

## Suggested citation and interpretation note

> An explicit identifier crosswalk to MODAVIS Pipe Organ Dataset 1.6.0 mapped
> all 119 historical acquisition-inventory entries to public virtual-instrument
> identities. Of these, 22 entries had accepted relationships to 21 distinct
> physical-organ identities. All ten VPOs used in the comparative spectral study
> had virtual-instrument identities, but none had an accepted physical-organ
> relationship in that release. These figures describe published identity-link
> coverage, not acquisition completeness or validation of the sampled organ state.

## Attribution

Source: MODAVIS Pipe Organ Dataset 1.6.0, public Navigator API and documented entity
URI policy. Original public responses, source URLs, source-row hashes and relation
hashes are retained for attribution and replay. Upstream metadata retains its
applicable terms; this package grants no new rights to source recordings.