Commit ·
69a3e24
1
Parent(s): ba7e790
Add reviewed physical-reference relationship layer (#29)
Browse files- Add reviewed physical-reference relationship layer (b7372af00fb899f178470619ad4d7eca1fcfd362)
- Archive_Specifications.md +23 -4
- Physical_References/CAPTURE_AND_INGEST_PROTOCOL.md +129 -0
- Physical_References/MANIFEST.json +40 -0
- Physical_References/SCHEMA.md +83 -0
- Physical_References/endpoint_references.csv +0 -0
- README.md +22 -1
- SCHEMA.md +21 -0
Archive_Specifications.md
CHANGED
|
@@ -17,13 +17,19 @@ post-processing history.
|
|
| 17 |
|
| 18 |
## Capture Configuration
|
| 19 |
|
| 20 |
-
The
|
| 21 |
|
| 22 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
- Capture resolution class: 2160p UHD (4K)
|
| 24 |
- Still-image size: 4000 × 3000
|
| 25 |
-
|
| 26 |
-
-
|
|
|
|
|
|
|
| 27 |
|
| 28 |
### Capture versus published encoding
|
| 29 |
|
|
@@ -36,6 +42,19 @@ landscape sources.
|
|
| 36 |
Do not infer a rotation, crop, or transcode solely from dimensions. Such a
|
| 37 |
transformation should be asserted only when supported by provenance evidence.
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
## Timing and Calibration Status
|
| 40 |
|
| 41 |
The current protocol documentation does not yet provide a complete calibrated
|
|
|
|
| 17 |
|
| 18 |
## Capture Configuration
|
| 19 |
|
| 20 |
+
The owner-confirmed default workflow since July 13, 2025 is:
|
| 21 |
|
| 22 |
+
- Device: Samsung Galaxy S23 Ultra, model `SM-S918U`
|
| 23 |
+
- Application: Framelapse 2; observed current version 11.7
|
| 24 |
+
- Logical camera: Rear (Main), 1.8x
|
| 25 |
+
- Nominal capture interval: 0.2 seconds
|
| 26 |
+
- Capture orientation intent: Landscape
|
| 27 |
- Capture resolution class: 2160p UHD (4K)
|
| 28 |
- Still-image size: 4000 × 3000
|
| 29 |
+
|
| 30 |
+
AE/WB lock, autofocus, and stabilization are owner-reported for certainly the
|
| 31 |
+
majority of this period, not confirmed as universal per-record facts. Deviations
|
| 32 |
+
must be recorded prospectively and must not be filled retrospectively by guess.
|
| 33 |
|
| 34 |
### Capture versus published encoding
|
| 35 |
|
|
|
|
| 42 |
Do not infer a rotation, crop, or transcode solely from dimensions. Such a
|
| 43 |
transformation should be asserted only when supported by provenance evidence.
|
| 44 |
|
| 45 |
+
### Acquisition evidence and canonical assembly
|
| 46 |
+
|
| 47 |
+
Framelapse source fragments are retained unchanged as acquisition evidence. One
|
| 48 |
+
verified chronological InShot assembly remains the canonical media object for
|
| 49 |
+
each drawing record. Dynamic browser previews and physical-reference photographs
|
| 50 |
+
are noncanonical access or evidence layers linked through persistent identity.
|
| 51 |
+
|
| 52 |
+
A 2160 × 3840 portrait H.264, 30-fps InShot processing profile is documented as
|
| 53 |
+
a candidate. It does not become a canonical protocol boundary until one complete
|
| 54 |
+
natural drawing passes fragment-order, join, duration, stream, hash, and manifest
|
| 55 |
+
verification and receives explicit creator approval. Existing 60-fps canonical
|
| 56 |
+
records remain unchanged.
|
| 57 |
+
|
| 58 |
## Timing and Calibration Status
|
| 59 |
|
| 60 |
The current protocol documentation does not yet provide a complete calibrated
|
Physical_References/CAPTURE_AND_INGEST_PROTOCOL.md
ADDED
|
@@ -0,0 +1,129 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Candidate Capture and Ingest Protocol
|
| 2 |
+
|
| 3 |
+
Status: documented candidate; requires an explicitly approved first natural
|
| 4 |
+
record before becoming a canonical processing-profile boundary.
|
| 5 |
+
|
| 6 |
+
## Authority model
|
| 7 |
+
|
| 8 |
+
1. **Framelapse source fragments** are acquisition evidence. Preserve them
|
| 9 |
+
unchanged with original filename, byte size, duration, stream properties,
|
| 10 |
+
order, and SHA-256.
|
| 11 |
+
2. **One chronological InShot assembly** remains the canonical public media
|
| 12 |
+
object for a drawing record.
|
| 13 |
+
3. **Dynamic browser previews** are noncanonical access derivatives bound to
|
| 14 |
+
`record_id`, integrity identity, and a pinned source revision.
|
| 15 |
+
4. **Drawing-reference photographs** are separate noncanonical physical-state
|
| 16 |
+
evidence linked through the sidecar only after review.
|
| 17 |
+
|
| 18 |
+
Making every Framelapse fragment a canonical row would change the archive's
|
| 19 |
+
unit of observation from one drawing record to an editing fragment. This
|
| 20 |
+
candidate rejects that change.
|
| 21 |
+
|
| 22 |
+
## Current documented acquisition profile
|
| 23 |
+
|
| 24 |
+
The following is supported for the current workflow but must not be silently
|
| 25 |
+
applied to every historical file:
|
| 26 |
+
|
| 27 |
+
- device: Samsung Galaxy S23 Ultra, model `SM-S918U`
|
| 28 |
+
- current documented environment: Android 14 / One UI 6.1
|
| 29 |
+
- application: Framelapse 2, observed version 11.7
|
| 30 |
+
- logical camera: Rear (Main), owner-confirmed at 1.8x since 2025-07-13
|
| 31 |
+
- interval: 0.2 seconds; nominal capture metadata: 5 captures per real-world second
|
| 32 |
+
- current controls: AE/WB lock, autofocus, and stabilization
|
| 33 |
+
- source class: 2160p UHD/4K HEVC; tested source files have no audio stream
|
| 34 |
+
|
| 35 |
+
Retrospective boundary: AE/WB lock, autofocus, and stabilization are
|
| 36 |
+
owner-reported for certainly the majority since 2025-07-13, not confirmed for
|
| 37 |
+
every record. Exact physical sensor/lens, camera-to-paper distance, angle,
|
| 38 |
+
lighting geometry, and spatial calibration remain unresolved.
|
| 39 |
+
|
| 40 |
+
The nominal five-captures-per-second tag is not proof of a perfectly calibrated
|
| 41 |
+
time base, zero jitter, or zero dropped capture moments.
|
| 42 |
+
|
| 43 |
+
## Candidate processing profile
|
| 44 |
+
|
| 45 |
+
- application: InShot, observed version 2.222.1548
|
| 46 |
+
- assemble source fragments in verified chronological order
|
| 47 |
+
- export 2160 x 3840 portrait MP4, H.264 High, 30-fps encoded playback
|
| 48 |
+
- no intentional trimming, speed change, crop, filter, transition, inserted
|
| 49 |
+
frame, or reordering
|
| 50 |
+
- no soundtrack or intentional audio; verify the stream because tested InShot
|
| 51 |
+
exports add an encoded silent AAC track
|
| 52 |
+
- preserve every Framelapse source fragment unchanged
|
| 53 |
+
- never re-encode the existing 60-fps corpus simply for uniformity
|
| 54 |
+
|
| 55 |
+
Controlled testing found that the 30-fps InShot output removed systematic
|
| 56 |
+
duplicate-frame pairs present in the tested 60-fps export and reduced assembled
|
| 57 |
+
file size by about half. It did not increase the underlying real-world sampling
|
| 58 |
+
rate. This remains a candidate until a naturally produced drawing passes the
|
| 59 |
+
full gate below.
|
| 60 |
+
|
| 61 |
+
## Daily ritual: create, record, preserve, assemble, verify, ingest, repeat
|
| 62 |
+
|
| 63 |
+
### Create
|
| 64 |
+
|
| 65 |
+
Make the drawing. Record only deviations that would otherwise be forgotten;
|
| 66 |
+
do not interrupt the practice to classify the archive.
|
| 67 |
+
|
| 68 |
+
### Record
|
| 69 |
+
|
| 70 |
+
Use the documented Framelapse profile. Multiple chronological fragments are
|
| 71 |
+
acceptable when needed.
|
| 72 |
+
|
| 73 |
+
### Preserve
|
| 74 |
+
|
| 75 |
+
Retain every fragment unchanged before assembly. Never rename or transcode the
|
| 76 |
+
only copy. Record original filename, bytes, duration, streams, and SHA-256.
|
| 77 |
+
|
| 78 |
+
### Assemble
|
| 79 |
+
|
| 80 |
+
Place fragments in observed chronological order in InShot. Apply the candidate
|
| 81 |
+
4K/30-fps profile without presentation edits.
|
| 82 |
+
|
| 83 |
+
### Verify
|
| 84 |
+
|
| 85 |
+
Confirm all expected fragments occur exactly once, order is correct, joins
|
| 86 |
+
omit nothing and add no transition, output duration agrees with the source sum,
|
| 87 |
+
output properties match the candidate profile, and source fragments remain
|
| 88 |
+
unchanged.
|
| 89 |
+
|
| 90 |
+
### Manifest
|
| 91 |
+
|
| 92 |
+
Write one assembly manifest with `record_id`, ordered fragment identities,
|
| 93 |
+
source/output hashes, capture and processing settings, software versions,
|
| 94 |
+
evidence state, and explicit nulls for unresolved variables.
|
| 95 |
+
|
| 96 |
+
### Ingest
|
| 97 |
+
|
| 98 |
+
Upload only the reviewed InShot assembly as canonical record media with the
|
| 99 |
+
normal 29-field metadata row. Keep detailed fragment relationships in a
|
| 100 |
+
separate manifest. Ingest physical-reference photographs only as noncanonical
|
| 101 |
+
linked media with their own hashes and privacy review.
|
| 102 |
+
|
| 103 |
+
### Derive
|
| 104 |
+
|
| 105 |
+
Generate previews only from the reviewed canonical record and bind them to
|
| 106 |
+
`record_id`, an integrity reference, and the pinned source revision.
|
| 107 |
+
|
| 108 |
+
### Repeat
|
| 109 |
+
|
| 110 |
+
Begin the next drawing. Version the protocol only when the change is intentional
|
| 111 |
+
and documented.
|
| 112 |
+
|
| 113 |
+
## Promotion gate
|
| 114 |
+
|
| 115 |
+
The 4K/30-fps candidate becomes canonical only after:
|
| 116 |
+
|
| 117 |
+
1. one complete drawing is naturally produced under this workflow;
|
| 118 |
+
2. its complete fragment set and assembly pass verification;
|
| 119 |
+
3. a finalized manifest binds all sources to the assembly;
|
| 120 |
+
4. the creator explicitly approves the result and boundary date;
|
| 121 |
+
5. archive specifications and the README are updated;
|
| 122 |
+
6. existing 60-fps records remain documented and unchanged.
|
| 123 |
+
|
| 124 |
+
## Privacy gate
|
| 125 |
+
|
| 126 |
+
Public derivatives must exclude phone numbers, serial numbers, IMEIs, precise
|
| 127 |
+
location metadata, and other private device/account identifiers. Original setup
|
| 128 |
+
photographs with GPS metadata are not publication-safe without metadata removal
|
| 129 |
+
and visual review.
|
Physical_References/MANIFEST.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"layer_name": "physical-reference-relationships-v0.1",
|
| 3 |
+
"release_state": "reviewed_relationship_table_reference_media_not_bundled",
|
| 4 |
+
"created_date": "2026-08-26",
|
| 5 |
+
"dataset_id": "maxwellinked/time-lapse-artifacts",
|
| 6 |
+
"source_hub_revision": "ba7e7903901767f2d5a4367d7ae22c2546366d6d",
|
| 7 |
+
"canonical_metadata_path": "metadata.csv",
|
| 8 |
+
"sidecar_path": "Physical_References/endpoint_references.csv",
|
| 9 |
+
"sidecar_sha256": "f0071876769ed4d90ba577ce76535cdc62b44ab27b5a1c575c2542eb60b62152",
|
| 10 |
+
"schema_path": "Physical_References/SCHEMA.md",
|
| 11 |
+
"capture_protocol_path": "Physical_References/CAPTURE_AND_INGEST_PROTOCOL.md",
|
| 12 |
+
"canonical_schema_field_count": 29,
|
| 13 |
+
"record_identity": "record_id",
|
| 14 |
+
"reference_identity": "reference_id",
|
| 15 |
+
"reference_media_bundled": false,
|
| 16 |
+
"internal_review_ledgers_bundled": false,
|
| 17 |
+
"counts": {
|
| 18 |
+
"standard_records_at_source_revision": 363,
|
| 19 |
+
"owner_confirmed_reference_links": 309,
|
| 20 |
+
"standard_records_without_current_reference_link": 54,
|
| 21 |
+
"relationship_types": {
|
| 22 |
+
"endpoint_photograph": 52,
|
| 23 |
+
"longitudinal_continuation": 1,
|
| 24 |
+
"owner_confirmed_record_reference": 232,
|
| 25 |
+
"two_sided_work_reference": 24
|
| 26 |
+
},
|
| 27 |
+
"reference_image_hash_state": {
|
| 28 |
+
"missing": 13,
|
| 29 |
+
"present": 296
|
| 30 |
+
}
|
| 31 |
+
},
|
| 32 |
+
"review_basis": "Owner-confirmed visual audit against canonical videos and retained physical-reference photographs; supporting review ledgers remain local and are not public dataset records.",
|
| 33 |
+
"limitations": [
|
| 34 |
+
"Reference photographs are not bundled in version 0.1 and are not downloadable from the Hub.",
|
| 35 |
+
"The table does not create frame labels, calibrated trajectories, or an official benchmark split.",
|
| 36 |
+
"Encoded playback rate is not real drawing time.",
|
| 37 |
+
"Exact temporal endpoint equivalence is asserted only where endpoint_claim_state supports it.",
|
| 38 |
+
"The corpus represents one creator and must not be generalized broadly without external validation."
|
| 39 |
+
]
|
| 40 |
+
}
|
Physical_References/SCHEMA.md
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Physical Reference Sidecar Schema
|
| 2 |
+
|
| 3 |
+
## Contract
|
| 4 |
+
|
| 5 |
+
`Physical_References/endpoint_references.csv` is a derived relational sidecar. It is not a third
|
| 6 |
+
canonical media index and it does not modify the 29-field video metadata
|
| 7 |
+
contract. Join it to `metadata.csv` on `record_id` after pinning both artifacts
|
| 8 |
+
to the same repository revision.
|
| 9 |
+
|
| 10 |
+
Each row represents one active, owner-confirmed relationship between a
|
| 11 |
+
canonical Standard video record and a photographed state or side of a physical
|
| 12 |
+
work. Version 0.1 has one row per linked `record_id`.
|
| 13 |
+
|
| 14 |
+
## Identity fields
|
| 15 |
+
|
| 16 |
+
| Field | Type | Meaning |
|
| 17 |
+
| --- | --- | --- |
|
| 18 |
+
| `reference_id` | string | Stable derived identity: `tlar-` plus 16 lowercase hex characters. |
|
| 19 |
+
| `record_id` | string | Persistent canonical video identity. |
|
| 20 |
+
| `canonical_file_name` | string | Repository-relative video path at `source_hub_revision`. |
|
| 21 |
+
| `source_hub_revision` | 40-char hex | Exact Hub revision used for the join. |
|
| 22 |
+
|
| 23 |
+
## Record context
|
| 24 |
+
|
| 25 |
+
`record_date`, `record_time`, `dimensions`, `tool`, `medium`, and `support` are
|
| 26 |
+
copied from the pinned canonical metadata for review convenience. They must not
|
| 27 |
+
be independently edited in the sidecar. Blank time remains unknown.
|
| 28 |
+
|
| 29 |
+
## Reference fields
|
| 30 |
+
|
| 31 |
+
| Field | Type | Meaning |
|
| 32 |
+
| --- | --- | --- |
|
| 33 |
+
| `reference_image_filename` | string | Source photograph filename; descriptive, not identity. |
|
| 34 |
+
| `reference_image_number` | string | Retained source number token. |
|
| 35 |
+
| `reference_image_role` | enum-like string | Literal role from the reviewed owner ledger. |
|
| 36 |
+
| `paired_image_filename` | string or blank | Associated back/opposite-side evidence when present. |
|
| 37 |
+
| `reference_image_sha256` | hex or blank | SHA-256 only when found in a current source ledger. |
|
| 38 |
+
| `paired_image_sha256` | hex or blank | SHA-256 only when found in a current source ledger. |
|
| 39 |
+
| `hash_state` | string | Explicit complete, missing, or conflict state. |
|
| 40 |
+
| `source_photo_collection` | semicolon list | Local source-ledger origin, not a public URL. |
|
| 41 |
+
|
| 42 |
+
## Relationship fields
|
| 43 |
+
|
| 44 |
+
| Field | Meaning |
|
| 45 |
+
| --- | --- |
|
| 46 |
+
| `relationship_type` | Conservative relationship category derived from explicit evidence. |
|
| 47 |
+
| `surface_scope` | Front, side A/B, unresolved side, or later front state. |
|
| 48 |
+
| `temporal_scope` | Whether endpoint, longitudinal, or temporally unspecified. |
|
| 49 |
+
| `endpoint_claim_state` | Explicitly states whether an endpoint claim is supported. |
|
| 50 |
+
| `pairing_state` | State of the associated paired/back image. |
|
| 51 |
+
| `owner_confirmation_state` | Owner-confirmed record link in every current row. |
|
| 52 |
+
| `evidence_state` | Lossless semicolon list copied from the active mapping ledger. |
|
| 53 |
+
| `owner_evidence` | Human-readable creator confirmation. |
|
| 54 |
+
| `paired_image_evidence` | Human-readable reverse/back qualification. |
|
| 55 |
+
|
| 56 |
+
Allowed `relationship_type` values in this candidate:
|
| 57 |
+
|
| 58 |
+
- `endpoint_photograph`
|
| 59 |
+
- `endpoint_side_reference`
|
| 60 |
+
- `longitudinal_continuation`
|
| 61 |
+
- `owner_confirmed_record_reference`
|
| 62 |
+
- `two_sided_work_reference`
|
| 63 |
+
|
| 64 |
+
`longitudinal_continuation` must never be used as exact endpoint supervision.
|
| 65 |
+
Two-sided relationships must not be collapsed into one assumed visual surface.
|
| 66 |
+
|
| 67 |
+
## Media availability
|
| 68 |
+
|
| 69 |
+
`reference_media_availability` is `local_evidence_only_not_bundled_v0.1`.
|
| 70 |
+
Photographs are not downloadable in this release. The two proposed-path
|
| 71 |
+
columns are intentionally blank and must not be interpreted as Hub paths.
|
| 72 |
+
|
| 73 |
+
## Missing values
|
| 74 |
+
|
| 75 |
+
Blank means unknown, not documented, unavailable in current ledgers, or not
|
| 76 |
+
applicable. It never means zero or false. Do not fill missing hashes, timestamps,
|
| 77 |
+
side roles, or temporal states by visual guess.
|
| 78 |
+
|
| 79 |
+
## ML boundary
|
| 80 |
+
|
| 81 |
+
`machine_use_scope` and `caution` describe supported selection/alignment uses.
|
| 82 |
+
They are not task labels, training splits, performance claims, or permission to
|
| 83 |
+
treat encoded video time as physical drawing time.
|
Physical_References/endpoint_references.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
README.md
CHANGED
|
@@ -484,10 +484,31 @@ guaranteed. See `SCHEMA.md`.
|
|
| 484 |
`encoded_frame_rate` is encoded playback rate. It is not the original temporal
|
| 485 |
sampling rate of the drawing process.
|
| 486 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 487 |
## Acquisition and Stored-Video Orientation
|
| 488 |
|
| 489 |
The acquisition protocol effective July 13, 2025 is documented in
|
| 490 |
-
`Archive_Specifications.md`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 491 |
|
| 492 |
The protocol's capture orientation and a file's stored pixel orientation are
|
| 493 |
different concepts. The `orientation`, `width_px`, and `height_px` metadata
|
|
|
|
| 484 |
`encoded_frame_rate` is encoded playback rate. It is not the original temporal
|
| 485 |
sampling rate of the drawing process.
|
| 486 |
|
| 487 |
+
### Physical-reference relationship layer
|
| 488 |
+
|
| 489 |
+
[`Physical_References/endpoint_references.csv`](Physical_References/endpoint_references.csv)
|
| 490 |
+
links 309 Standard video records to owner-confirmed photographs of their
|
| 491 |
+
physical drawings through persistent `record_id` values and the pinned source
|
| 492 |
+
revision. It is a separate derived relationship table and does not alter the
|
| 493 |
+
29-field canonical video index.
|
| 494 |
+
|
| 495 |
+
Version 0.1 records 52 confirmed endpoint photographs, one longitudinal
|
| 496 |
+
continuation, 232 owner-confirmed record references whose exact endpoint state
|
| 497 |
+
is not separately asserted, and 24 two-sided references. The photographs are
|
| 498 |
+
not bundled in this release. Their reviewed filenames and available hashes
|
| 499 |
+
preserve relationship evidence without implying that the media are publicly
|
| 500 |
+
downloadable. See the [sidecar schema](Physical_References/SCHEMA.md) and
|
| 501 |
+
[manifest](Physical_References/MANIFEST.json).
|
| 502 |
+
|
| 503 |
## Acquisition and Stored-Video Orientation
|
| 504 |
|
| 505 |
The acquisition protocol effective July 13, 2025 is documented in
|
| 506 |
+
`Archive_Specifications.md`. Framelapse source fragments remain unchanged
|
| 507 |
+
acquisition evidence, while one verified chronological InShot assembly remains
|
| 508 |
+
the canonical media object for each drawing record. A documented 4K/30-fps
|
| 509 |
+
processing profile remains a candidate and does not retroactively redefine
|
| 510 |
+
existing records. See the
|
| 511 |
+
[capture and ingest protocol](Physical_References/CAPTURE_AND_INGEST_PROTOCOL.md).
|
| 512 |
|
| 513 |
The protocol's capture orientation and a file's stored pixel orientation are
|
| 514 |
different concepts. The `orientation`, `width_px`, and `height_px` metadata
|
SCHEMA.md
CHANGED
|
@@ -179,6 +179,27 @@ Observed evidence methods include:
|
|
| 179 |
|
| 180 |
New enumerated values must be documented here before use.
|
| 181 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182 |
## Missing-Value Rules
|
| 183 |
|
| 184 |
- Blank means unknown, not documented, or not applicable according to the field
|
|
|
|
| 179 |
|
| 180 |
New enumerated values must be documented here before use.
|
| 181 |
|
| 182 |
+
## Physical-reference sidecar
|
| 183 |
+
|
| 184 |
+
`Physical_References/endpoint_references.csv` is an optional derived
|
| 185 |
+
relationship table joined to the canonical Standard index on `record_id`. It
|
| 186 |
+
must declare the exact `source_hub_revision`; it does not determine canonical
|
| 187 |
+
video status or duplicate editable video metadata.
|
| 188 |
+
|
| 189 |
+
Each row has a stable `reference_id`, reviewed photograph filename and role,
|
| 190 |
+
explicit surface and temporal scope, evidence state, hash state, and machine-use
|
| 191 |
+
caution. Blank values mean unknown or not documented. Version 0.1 does not
|
| 192 |
+
bundle the reference photographs, and blank proposed-path fields must not be
|
| 193 |
+
interpreted as public media locations.
|
| 194 |
+
|
| 195 |
+
`longitudinal_continuation` identifies a later or otherwise changed physical
|
| 196 |
+
state and is not an exact endpoint target. Two-sided references preserve
|
| 197 |
+
side-specific evidence and do not imply one shared visual target. Validators
|
| 198 |
+
must require unique identities, canonical filename agreement at the pinned
|
| 199 |
+
revision, allowed relationship values, explicit owner-confirmation state, and
|
| 200 |
+
valid SHA-256 syntax whenever a hash is present. The complete contract is in
|
| 201 |
+
[`Physical_References/SCHEMA.md`](Physical_References/SCHEMA.md).
|
| 202 |
+
|
| 203 |
## Missing-Value Rules
|
| 204 |
|
| 205 |
- Blank means unknown, not documented, or not applicable according to the field
|