# OrgRec VPO Collection ## Provisional inventory and source availability audit Author: Dominik Ukolov, ORCID https://orcid.org/0000-0002-7904-3892 Version: 0.1.0 Prepared: 10 September 2026 Reserved DOI: **10.5281/zenodo.22693457** Status: **Unpublished draft. Not a deposit of the sample payloads.** This reference package documents the available evidence for the historical VPO discovery collection used during development of OrgRec. It preserves the distinction between catalog discovery, acquisition-log evidence, filesystem inspection, and the ten VPO entities used in the comparative stop-timbre study. It does not establish complete collection membership or byte-equivalent reconstruction of the historical collection. ## Start here - **VPO_Collection_Reference.pdf**: concise reference document with the complete retained collection-evidence listing and planned extensions. - **VPO_Collection_Reference.html**: searchable version with provider links. - **All_VPO_Catalog.html**: all 3,600 retained discovery-catalog entries, including products, definitions, variants and components. These are not 3,600 acquired VPOs. - **Native_Definition_Index.html**: all 235 historically inspected definition files with logical package-relative paths and formats. - **AVAILABILITY_REPORT.md**: dated summary of endpoint checks and exceptions. - **DATA_DICTIONARY.md**: machine-readable data definitions and interpretation. - **MODAVIS_CROSSWALK.md** and **modavis-crosswalk.json**: exact catalog-ID mappings to persistent identities in MODAVIS POD 1.6.0, with scope-specific organ-link counts, frozen public responses and offline replay code. - **CITATION.cff** and **CITATION.bib**: draft citation metadata. The DOI is reserved, not registered as a published record; update release status at publication. ## Evidence and counts The collection listing contains 123 evidence records: all 119 retained inventory rows, two additional identifiers seen in native-definition parent directories, one reported incomplete acquisition, and one uncataloged SoundFont package. The two directory labels are not validated independent instrument attributions. The provisional identifier UNASSIGNED-001 is local to this reference, not an original VPO catalog identifier. The inventory's 105 distinct sampled-instrument name strings are not a verified count of complete playable sample sets. Its recorded successful acquisition status includes one website GIF and six demonstration MP3s. These rows are retained and explicitly qualified. Other rows are not automatically treated as complete simply because their paths resemble archives or extracted directories. Known legacy grouped records contain attribution errors. Missing exact archive paths can result from removal after extraction and do not prove missing payloads. The 31 August 2026 inspection reports 551,769 regular files and 582,098,581,740 logical bytes. The informal approximately 591 GB collection label, the roughly 591 GiB acquisition notes and this inspection total remain unreconciled. No new raw-payload inventory, full-file checksum pass or full recovery test was performed for this release. Source input fingerprints identify the retained reports; they are not hashes of the complete VPO payload collection. The catalog contains 646 dataset/product entries, 223 derived instruments or organ definitions, 46 edition/platform variants, 329 configuration variants and 2,356 stop/rank components. A catalog row is not necessarily independently distributed, independently recorded, acquired, or suitable as a statistical observation. ## Repeat the availability audit Requires Python 3.10 or newer; the checker uses only the standard library. From this directory, choose a new output filename to preserve the archived audit: ```sh python3 check_availability.py --registry source-registry.json --output availability-new.json ``` The registry contains all 197 unique primary source URLs from the retained catalog and ten reviewed direct archive routes from the analyzed subset. Secondary URLs are retained in catalog.json but not included in this audit. A primary source URL can be a generic provider page. Its successful response does not confirm a particular product edition or locate every download component. The checker issues HEAD requests to source pages and HEAD plus a bounded range request to known direct archives. It reads at most 4,096 response-body bytes per range request at application level, closes the response even if the server ignores Range, and never transfers complete large packages. HTTP headers and protocol/network buffering are additional to this application-read count. It uses no account credentials, bypasses no access controls, and does not accept source terms on a user's behalf. A site blocking HEAD is recorded as an indeterminate or restricted result rather than assumed unavailable. Current checks are time-specific and can be affected by rate limits or network conditions. Interpretation: - Source-page headers reachable: the endpoint responded, not a proven download. - Limited archive probe passed: recognized ZIP/RAR/7z prefix, not full integrity. - Rate limited: retry at a later time; do not infer removal. - Not found: HTTP 404 or 410 at the checked endpoint; other lawful routes may exist. - Access restricted or blocked: may require authentication or manual review. - Indeterminate network error: transport failure; availability is unresolved. - Unconfirmed content or HTML in place of an archive: not validated as a package. Reported size equality is a change detector, not cryptographic identity. The audit does not establish complete multipart retrieval or whole-corpus equivalence. ## Rebuild the reference documents Install reportlab in a separate Python environment and provide the directory containing DejaVuSans.ttf and DejaVuSans-Bold.ttf: ```sh python3 render_reference.py --package . --font-dir /path/to/dejavu-fonts --doi 10.5281/zenodo.22693457 python3 -m unittest test_availability.py python3 verify_reference.py ``` Render into a working copy, not the immutable archived reference. The renderer uses the frozen JSON evidence and the dated audit; no network retrieval is required. PDF typography and embedded metadata can differ across library versions. The package checksums certify the deposited files, not a promise of byte-identical PDF regeneration on every platform. Environment versions for this build are recorded in build-environment.json. The existing analysis-specific acquisition and reproduction implementation is separate: https://github.com/modavis-project/orgrec (Tools/vpo_reproduce.py). Cite the exact software and study artifact versions used. OrgRec 0.3.0 has DOI https://doi.org/10.5281/zenodo.22216026. This reference package does not bundle all inputs needed to reproduce the comparative study. ## Planned extensions The MODAVIS identity crosswalk is already integrated: 119/119 inventory entries have public identities, and 22 link to 21 distinct organs. The broader listing has 122/123 identities and 23 linked entries referring to 22 organs. All ten analyzed VPOs have identities but no accepted organ links in POD 1.6.0. See MODAVIS_CROSSWALK.md for denominators and interpretation. These counts do not change the acquisition evidence or HTTP audit findings. 1. Reconcile preserved source payloads with original acquisition logs and all surviving metadata; resolve ambiguous identities and duplicate editions. 2. Establish a full per-file SHA-256 manifest, exact logical byte count and explicit scope for incomplete transfers, auxiliary files and acquisition residue. 3. Record exact source versions, archive hashes where available, all multipart dependencies, access conditions and provenance confidence. 4. Extend availability checks to resolved download components and alternative authorized routes, with a documented review of redistribution rights. 5. Provide a materializer using authorized external sources or locally supplied archives, with strict completeness checks and explicit non-equivalent variant reports. 6. Release new versioned evidence and validation results. Additional spectral analyses and native-format adapters remain separate scientific extensions. These are intended extensions, not capabilities already delivered by this release. Full public-source reconstruction may remain impossible for unavailable or access-restricted exact editions. Local reconstruction and public reacquisition must be evaluated separately. ## Suggested citation and reference text While the record is an unpublished draft: > A provisional reference package documents the retained VPO collection evidence, > discovery catalog and timestamped source-availability checks. Its DOI has been > reserved, but the record has not yet been published. The package does not > redistribute source audio or establish whole-collection byte equivalence. After publication, cite the released version and its version-specific DOI. Do not describe an endpoint probe as verified complete downloading or the catalog as a collection of independent organs. ## Rights and publication review The proposed licence for the authored documentation and compilation is CC BY 4.0; the included code is Apache-2.0. Third-party audio, images, instrument definitions and private acquisition logs are not included. The historical licence labels in the catalog are discovery metadata, not permissions granted by this record. Provider names, factual metadata and source references retain attribution. Before publication, confirm creators, affiliations if desired, licence scope, description and release date. No publication action is included in this package. ## References - Wilkinson, M. D., et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3, 160018. https://doi.org/10.1038/sdata.2016.18. These principles motivate persistent identification, provenance and machine-readable evidence; no full FAIR compliance claim is made here. - Zenodo. DOI reservation. https://help.zenodo.org/docs/deposit/describe-records/reserve-doi/ - Zenodo. Manage versions. https://help.zenodo.org/docs/deposit/manage-versions/ - Zenodo. General policies. https://about.zenodo.org/policies/ Online guidance consulted 10 September 2026. Historical input names and SHA-256 fingerprints are in historical-summary.json. The original private reports are not redistributed verbatim. Source-specific references are in catalog.json and source-registry.json.