synapse-cdm 2.1.2__tar.gz → 2.2.0__tar.gz

{synapse_cdm-2.1.2/synapse_cdm.egg-info → synapse_cdm-2.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: synapse-cdm
-Version: 2.1.2
+Version: 2.2.0
 Summary: SynapseCommand Canonical Data Model: the four canonical objects, the adapter SDK, and the validation harness
 Author: Decent Cybersecurity s.r.o.
 License-Expression: Apache-2.0
@@ -33,6 +33,8 @@ Requires-Dist: jsonschema>=4.0
 Provides-Extra: test
 Requires-Dist: pytest>=8.0; extra == "test"
 Requires-Dist: rdflib>=7.0; extra == "test"
+Provides-Extra: lint
+Requires-Dist: ruff==0.16.6; extra == "lint"
 Dynamic: license-file
 
 # `synapse_cdm` — the Canonical Data Model and adapter framework
@@ -78,7 +80,7 @@ With one, an adapter is a thin translator and nothing else.
 pip install synapse-cdm
 
 python -m synapse_cdm.harness --adapter pntmap        # replays the fixtures that came with it
-python -m synapse_cdm.schemas --out ./schemas         # writes the six JSON Schemas, anywhere
+python -m synapse_cdm.schemas --out ./schemas         # writes the eight JSON Schemas, anywhere
 
 python -m synapse_cdm.harness --list-adapters         # the names --adapter takes
 ```
@@ -257,11 +259,14 @@ golden file, not by a failing test.
 
 **7. Add the tests.** Copy the shape of `tests/test_cdm_pntmap_adapter.py`: one test per claim
 in your adapter's docstring. If you are bidirectional, the harness already round-trips you —
-declare `direction = "bidirectional"`, override `from_cdm()`, and the `roundtrip` column checks
-that no source value goes missing on the way out. It compares values, not bytes: a byte-equal
-round trip is neither achievable nor the point (key order changes, omitted optional fields come
-back explicit, XML attribute order is arbitrary). An adapter emitting XML or USMTF gets `SKIP`
-there and must ship its own round-trip test.
+declare `direction = "bidirectional"`, override `from_cdm()`, and the `roundtrip` column judges
+what comes back. A JSON emitter is compared by values, not bytes (key order changes, omitted
+optional fields come back explicit). Anything else is compared under the tolerance the class
+declares in `ROUNDTRIP_TOLERANCE`, printed with every report: `bytes`, the default, is octet
+equality against the byte fixture; `values`, for XML, re-ingests what was emitted and requires no
+source value missing from the parsed twin, with `ROUNDTRIP_TRANSFORMS` naming what egress
+legitimately re-stamps. Your own round-trip test in `tests/` is your statement of the claim; the
+column is the suite's.
 
 ## The harness
 
@@ -308,8 +313,8 @@ in that case, because the shape of a report is itself a claim that fixtures were
 **The fixture directory is not always the adapter's name, and you no longer have to know that.**
 `stanag4676` reads its fixtures from `fixtures/nits` — the adapter is named for a covering
 document and the directory for the bytes it holds — and pointing `--fixtures` at
-`fixtures/stanag4676` is the invocation that used to pass vacuously, because that directory holds
-only pinned standards. Each adapter now DECLARES its directory (`Adapter.fixture_dir`) and the
+`fixtures/stanag4676` is the invocation that used to pass vacuously, because that directory held
+only pinned standards (it has since been removed; the pins are in `fixtures/nits/spec/`). Each adapter now DECLARES its directory (`Adapter.fixture_dir`) and the
 harness resolves it through `importlib.resources`, so omitting `--fixtures` is always right for a
 shipped adapter. `tests/test_cdm_harness.py` holds the same map written out by hand and requires
 the two to agree, so a new adapter cannot join the roster without one.
@@ -355,25 +360,21 @@ So audit the model separately, against something outside the implementation —
 - `asterix_cat021`'s scale factors are the safe kind: each is a single stated LSB, checkable
   against the document without running anything.
 
-**A wheel-only consumer cannot run any round-trip proof for an adapter with a non-JSON egress,
-and the harness says so in a sentence that points where the wheel does not reach.** `roundtrip`
-reports `SKIP` when `from_cdm()` returns bytes it cannot compare structurally, and the SKIP text
-reads "the adapter must ship its own round-trip test in `tests/`". That instruction is correct and
-it is unreachable from a wheel: `tests/` is not packaged, so a consumer who installed from PyPI
-reads a pointer to a directory they do not have. It is not a defect in any adapter and not a defect
-in a release — it is the shape of the distribution, and it has been true of every version. What it
-costs is specific rather than general: `lossless` and the schema checks still run and still prove
-what they prove, so the floor a wheel-only consumer gets is **ingress** conformance, and egress
-byte-exactness is proved only in a clone. Every adapter that declares an egress direction is
-affected — eleven of the fourteen shipped adapters, every one of which emits something the check
-cannot parse as JSON, leaving only the three ingest-only adapters unaffected for a different
-reason.
-Two things would change it and neither is free: packaging the round-trip tests, which puts a test
-suite inside a runtime distribution; or giving the harness a comparison that works on the emitted
-bytes per format, which is the codec-level work each of those adapters already does in `tests/`.
-Recorded here rather than fixed in passing, and recorded as reach rather than as a count of one
-release: it is what a wheel-only conformance claim does NOT cover, and the person who needs to know
-is the one reading `20 passed, 0 failed` from an installed copy.
+**A wheel-only consumer CAN run the round-trip proof for every emitter, since 2026-09-16.** Until
+then `roundtrip` reported `SKIP` when `from_cdm()` returned bytes it could not compare
+structurally, and the SKIP text read "the adapter must ship its own round-trip test in `tests/`" —
+an instruction that was correct and unreachable from a wheel, because `tests/` is not packaged, so
+a consumer who installed from PyPI read a pointer to a directory they did not have. The floor a
+wheel-only consumer got was **ingress** conformance, and egress byte-exactness was proved only in a
+clone. Every adapter that declares an egress direction was affected — eleven of the fourteen
+shipped adapters, every one of which emits something the check could not parse as JSON. The second
+of the two repairs that paragraph named was taken: the harness compares the emitted bytes itself,
+under the tolerance each class declares in `ROUNDTRIP_TOLERANCE` — octet for octet for the nine
+binary and line-oriented codecs, and by re-ingest for the two XML emitters, whose format cannot
+promise octet order — and prints the declaration with every report. The tests in `tests/` still
+make the same claims; they are each adapter's own statement of it, and the column is the suite's.
+The person who needs to know is the one reading `20 passed, 0 failed` from an installed copy, and
+what that line covers now includes the way back.
 
 **The roster sweep is a manual protocol act, and prose counts are what it is for.** When an
 adapter joins the shipped roster, every document that restates how many adapters there are has to
@@ -808,15 +809,25 @@ packages/cdm/
   synapse_cdm/
     models.py       the four objects, Position, Kinematics, SourceRef, Integrity, payloads
     enums.py        closed vocabularies; UNKNOWN is a member, never a null
-    geo.py          GeoJSON Point/LineString/Polygon, [lon, lat], ring closure enforced
+    geo.py          GeoJSON Point/LineString/Polygon and their Multi* forms, [lon, lat], ring closure enforced
     times.py        one timestamp form, one injectable clock
     ids.py          derived stable identity (uuid5) and the id basis
-    version.py      SCHEMA_VERSION and the compatibility rule
+    version.py      the six version constants and the compatibility rule
     symbology.py    MIL-STD-2525D standard identity, CoT affiliation letters
     lossless.py     the never-drop rule as a computable check
+    canonical.py    the one serialisation (sort_keys, indent 2, trailing newline), written once
     adapter.py      the Adapter ABC, its class-definition-time gates, the registry
     schemas.py      JSON Schema export (+ --check for CI)
     harness.py      the adapter-agnostic validation harness
+    suite.py        Conformance Suite v2 and the `synapse` CLI: checks G–O over the harness's A–F
+    conformance.py  the SC-OES conformance tool, dimensions A–E
+    manifest.py     the adapter-manifest model
+    manifests.py    manifest generation for the shipped roster (+ --check for CI)
+    evidence.py     evidence records and badges; the one module allowed hashlib
+    oes.py          the SC-OES event block: type ids, ontology identifiers, effective intervals
+    oes_registry.py the packaged SC-OES registries, read offline, and their helper surface
+    release_notes.py release-note rendering for the release pipeline
+    registry/       the three packaged SC-OES registries, read as data
     adapters/       one module per external system (pntmap, tak, ais, adsb, …)
     fixtures/       synthetic payloads + golden outputs
 schemas/            published JSON Schema, generated — never hand-edited

{synapse_cdm-2.1.2 → synapse_cdm-2.2.0}/pyproject.toml

@@ -51,8 +51,10 @@ classifiers = [
     "Intended Audience :: Information Technology",
     # Enumerated rather than left to `Programming Language :: Python :: 3`, because the floor is
     # a real declaration here — `tests/test_cdm_version_floor.py` asserts the package PARSES at
-    # 3.11 — and the ceiling is what is actually exercised. 3.14 is the interpreter the suite
-    # runs on today; 3.11 is the floor that gate defends.
+    # 3.11 — and the ceiling is what is actually exercised. Since 2026-09-16 the suite runs on
+    # all four in `ci.yml`'s `suite` matrix, so each classifier is an executed reading rather
+    # than a declaration; the version-floor module still gates grammar at 3.11 on every
+    # interpreter, and 3.12 is the release interpreter `publish.yml` pins.
     "Programming Language :: Python :: 3",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
@@ -77,16 +79,24 @@ dependencies = [
 ]
 
 # `rdflib` is the RDF parser the ontology's own tests read `ontology/*.ttl` with, and it is a
-# TEST dependency and nothing else. SC-OES-SPEC-v2 §136 fixes all four conditions and this extra
+# TEST dependency and nothing else. ADR 0002 decision 6 and ADR 0010 decision 5 fix all four
+# conditions and this extra
 # is where the first of them lands: test/development only, licence confirmed compatible before it
 # landed, no RDF runtime dependency, the choice recorded in ADR 0002/0010 lineage. rdflib 7.6.0 is
 # BSD-3-Clause and its one non-optional dependency, pyparsing, is MIT — both permissive and both
 # compatible with Apache-2.0, which is the criterion `docs/adr/0010-...md` decision 5 states for a
 # test extra. Nothing under `synapse_cdm/` imports it: `tests/test_cdm_boundary.py` is the check,
-# and §102 is the reason — the runtime reads the GENERATED `registry/sc_oes/ontology_terms.json`
+# and `spec/sc-oes/README.md` ("Machine authorities") is the reason — the runtime reads the GENERATED `registry/sc_oes/ontology_terms.json`
 # and never parses Turtle.
 [project.optional-dependencies]
 test = ["pytest>=8.0", "rdflib>=7.0"]
+# The lint stage's ONE pin, added 2026-09-16. `ci.yml`'s `lint` job and `publish.yml`'s gate both
+# install this extra rather than typing a version of their own, so the tool version and the rule
+# set below are read from one file by both workflows and by a contributor following
+# CONTRIBUTING.md; a workflow line `ruff==` is what `tests/test_cdm_lint_stage.py` refuses. Pinned
+# exactly, not floored, because the rule set below was measured against this version and a newer
+# ruff can add a rule to `E9` or change what F821 sees.
+lint = ["ruff==0.16.6"]
 
 [project.urls]
 Homepage = "https://github.com/Decent-Cybersecurity/synapsecommand-public"
@@ -100,7 +110,8 @@ Changelog = "https://github.com/Decent-Cybersecurity/synapsecommand-public/blob/
 
 [project.scripts]
 # The entry points the README documents as `python -m ...`; every one of them stays importable
-# that way. `cdm-conformance` joined 2026-09-06 with the SC-OES conformance tool: §40 requires a
+# that way. `cdm-conformance` joined 2026-09-06 with the SC-OES conformance tool:
+# `spec/sc-oes/13-conformance.md` ("Exit codes") requires a
 # process status a CI system can branch on, and a console script is how this repository has
 # always spelled "runnable without knowing where the package is installed".
 cdm-harness = "synapse_cdm.harness:main"
@@ -138,6 +149,24 @@ synapse = "synapse_cdm.suite:main"
 # has no business making. A stage added red is a stage somebody switches off in the release it
 # first blocks; this one is added green and can only go red on something new.
 #
+# CORRECTED 2026-09-16: THE SENTENCE ABOVE DESCRIBED A SET THIS FILE DID NOT SELECT. Under ruff
+# 0.16.6, `select = ["E9"]` enables exactly one rule — `io-error` (E902), read off
+# `ruff check --show-settings` — a syntax error is reported regardless of selection, and "a name
+# that is not defined at all" is pyflakes' F821, which `E9` never included. So the stage as
+# configured caught syntax errors and unreadable files and never an undefined name, on a tag or
+# anywhere else. The set is now `E9,F821`, which is what the sentence meant: measured at 0
+# findings over `packages/cdm gates tests` on 2026-09-16, so it is added green, under the same
+# escape clause as the original narrowing. The rest of `F` (27 F401 on the same reading, all
+# `--fix`able) is still the documented limitation and still a ruling of its own. The version is
+# pinned once, in the `[lint]` extra above, and `ci.yml` now runs this set on every push.
+#
+# AND, LATER THE SAME DAY, THE F401s WERE REMOVED. The audit's helpers commit deleted every unused
+# import the reading above counted — 29 by then, two test modules having gained one each that
+# morning — and `ruff check --select F401` over the three roots reads 0. `select` is NOT widened
+# here: `tests/test_cdm_lint_stage.py` makes widening a ruling and an edit to both files, and the
+# readiness report's item 3 makes it a round of its own. What remains of `F` is 13 F841 and 8 F541,
+# measured with ruff 0.16.6 on 2026-09-16, none of which changes what the package does.
+#
 # There is no `[tool.mypy]`, and that is also F7.1: the package has never had a type config, and
 # adding one in a release-engineering round would put a first type-check run inside the pipeline
 # that publishes. P8 §57 records it as a limitation.
@@ -146,7 +175,7 @@ line-length = 100
 target-version = "py311"
 
 [tool.ruff.lint]
-select = ["E9"]
+select = ["E9", "F821"]
 
 [tool.setuptools.dynamic]
 version = { attr = "synapse_cdm.version.PACKAGE_VERSION" }
@@ -217,11 +246,29 @@ include = ["synapse_cdm*"]
 # shipped are named positively below.
 #   registry/**/*                      the machine-readable runtime artefacts of ADR 0004 —
 #                                      `registry/sc_oes/ontology_terms.json` today, the governed
-#                                      event registry beside it later. §17 requires them to ship
+#                                      event registry beside it later. `spec/sc-oes/03-event-types.md`
+#                                      ("The registry, and where it lives") requires them to ship
 #                                      in the wheel, load with `importlib.resources`, need no
 #                                      checkout and no network, and be read as DATA. A recursive
 #                                      glob for the same reason the fixtures one is recursive: the
 #                                      second artefact must not depend on somebody remembering.
+#
+# THE TWO LARGE DOCUMENTS SHIP WHOLE, AND THEIR SIZE IS RECORDED HERE RATHER THAN BOUNDED. Noted
+# 2026-09-16 by the repo-hygiene audit, which found the fact stated nowhere in the tree. The
+# `*.md` glob below carries `FORMAT_COVERAGE.md` — 1,887,457 bytes and 16,125 lines on that date,
+# the largest tracked file in this repository and nearly twice the next — and `MIGRATIONS.md`
+# (978,711 bytes, 12,037 lines) into every consumer's site-packages, and both grow on every round
+# that closes a park or records an arc: 93 historical blobs of the first (117 MB) and 140 of the
+# second (59 MB) delta-compress to a 10.37 MiB pack on the same date. They ship ON PURPOSE and the
+# shipping is load-bearing, not incidental: `tests/test_cdm_klv_framing.py` reads both from the
+# INSTALLED package through `synapse_cdm.__file__` — `gates/wheel_install.py` says why that module
+# is package-only — and the same gate refuses a wheel whose `.py`/`.md` count is not
+# `git ls-files synapse_cdm`'s, so a wheel that dropped either would not pass. The size is the
+# cost of the SHIPS row above (the protocol documents a partner reads travel with the code they
+# describe), and no ceiling is set on it: a ceiling would be one more figure typed in this file,
+# and a document that has grown too large to ship whole is a document to split, not to cap. The
+# figures are a reading and not a derivation — nothing re-derives them, and a later reader who
+# finds them stale has found the growth this note says to expect, not a defect.
 [tool.setuptools.package-data]
 synapse_cdm = [
     "fixtures/**/*",

{synapse_cdm-2.1.2 → synapse_cdm-2.2.0}/synapse_cdm/FORMAT_COVERAGE.md

@@ -744,7 +744,7 @@ reason is indistinguishable from "nobody thought about it".
 
 **Licensing posture, read 2026-09-07.** `fixtures/adsb/spec/adsb_terms.json` records both publishers, and both sell. `my.rtca.org/nc__store?search=DO-260B` answered **200** at **2026-09-07T20:54:47Z** and prices the document *"DO-260B - Electronic / Retail Price: $875.00"* and *"DO-260B - Hard Copy / Retail Price: $715.00"*; `www.rtca.org/content/electronic-license-agreement` answered **200** at **20:55:49Z** and reads *"YOU ACKNOWLEDGE THAT YOU HAVE PURCHASED ONLY THE RIGHT TO ACCESS, DOWNLOAD, AND PRINT ONE COPY … FOR YOUR PERSONAL USE AND NOT FOR RESALE, TRANSFER, OR DISTRIBUTION TO ANY THIRD PARTY."* `store.icao.int/en/annexes/annex-10?p=2` answered **200** at **20:54:25Z** and prices *"Annex 10 - Aeronautical Telecommunications - Volume IV- Surveillance Radar and Collision Avoidance Systems / Sixth Edition, August 2026 / USD 240.00"* — so the free-public-download branch that names ICAO is refuted by ICAO's own store, and no page read states a free route to an Annex. **Class: `LICENSED`**. Nothing was acquired: this is a reading of a publisher's page, not a pin.
 
-## Picogrid Legion Platform API v3 — ingest (specification only)
+## Picogrid Legion Platform API v3 — ingest
 
 Implemented by `adapters/legion.py` (**ingest only** — see the last section for why there is no
 `from_cdm()`). Every row below was written and reviewed as a specification BEFORE any code
@@ -4139,7 +4139,7 @@ corrections; a retraction-only `TrackSegment` that must become an `Event` and no
 `relTimeIncrement` that is not a whole number of milliseconds, so the parking rule is what the
 round trip depends on.
 
-## STANAG 4607 / AEDP-4607 — NATO Ground Moving Target Indicator Format (GMTIF), ingest
+## STANAG 4607 / AEDP-4607 — NATO Ground Moving Target Indicator Format (GMTIF), ingest and egress
 
 Implemented by `adapters/gmtif.py` (bidirectional), on the Annex C wire codec in
 `adapters/gmtif_codec.py`. Ingest translates one GMTI packet into an `Entity` and a `Track` for the
@@ -8073,7 +8073,7 @@ tag alone silently collapses two values into one. Structure is preserved verbati
 | `70` | Alternate Platform Name | None | `utf8` | `V` | `Entity.attributes` | `not yet` | Name of alternate platform connected to UAS |
 | `71` | Alternate Platform Heading | ° | `uint16` | `2` | `Entity.attributes` | `not yet` | Heading angle of alternate platform connected to UAS |
 | `72` | Event Start Time - UTC | µs | `uint64` | `8` | `Entity.attributes` | `not yet` | Start time of scene, project, event, mission, editing event, license, publication, etc. An instant, and the epoch it counts from is **MISB ST 0603.5's**, held and pinned since 2026-09-04 — settlement 3, settled. The row stays `not yet` on the scope contract and not on the epoch: the pinned stream does not carry tag 72 and §8.72 prints a worked example, so what it needs is a round scoped to it rather than a document. |
-| `73` | RVT Local Set | None | `set` | `V` | `Entity.attributes` | `stanag4609 1.0.0 · carried` | MISB ST 0806 RVT Local Set metadata items. A nested Local Set or Pack — the items inside it are **MISB ST 0806.4**'s, park 7. **PROMOTED 2026-09-06 BY THE PARK 7 ROUND.** The document is pinned and `adapters/klv_rvt_codec.py` reads all four of its tables — Table 8-1's RVT LS (21 tags), Table 8-2's POI LS (10), Table 8-3's AOI LS (10) and Table 8-4's User Defined LS (2) — with one recursive `decode_set`, because a subordinate set's Value is a bare run of triplets exactly as item 73's is. The elements ride in `Entity.attributes.rvt_local_set` **as the document names them and reach no canonical CDM field**, which is this row's own CDM field cell being obeyed rather than a shortfall: eight elements have an obvious home — the POI's latitude, longitude and altitude, the AOI's four corners and the RVT LS's own timestamp — and emitting a POI as a second `Entity` is a modelling decision no clause of either document makes. Stated at `klv_rvt_codec.CDM_MAPPING_NOT_TAKEN` and proposed in `rounds/reports/F.md`. **THE WITNESS IS `carried` AND NOT `mapped`, AND IT IS THE WEAKEST IN THIS ROW SET**: ST 0806.4 prints NO worked example — Figure 7-1 is a raster image — so the ground for reading these octets is item 48's second-document agreement (ST 0601.14a §8.73 and `ST 0806.4-06` state the same sixteen octets and CRC 17945) and not a printed example. Twelve internal cross-checks stand in place of examples and establish that the four tables were transcribed consistently; they establish nothing about any value map, each of which is stated once. |
+| `73` | RVT Local Set | None | `set` | `V` | `Entity.attributes` | `stanag4609 1.0.0 · carried` | MISB ST 0806 RVT Local Set metadata items. A nested Local Set or Pack — the items inside it are **MISB ST 0806.4**'s, park 7. **PROMOTED 2026-09-06 BY THE PARK 7 ROUND.** The document is pinned and `adapters/klv_rvt_codec.py` reads all four of its tables — Table 8-1's RVT LS (21 tags), Table 8-2's POI LS (10), Table 8-3's AOI LS (10) and Table 8-4's User Defined LS (2) — with one recursive `decode_set`, because a subordinate set's Value is a bare run of triplets exactly as item 73's is. The elements ride in `Entity.attributes.rvt_local_set` **as the document names them and reach no canonical CDM field**, which is this row's own CDM field cell being obeyed rather than a shortfall: eight elements have an obvious home — the POI's latitude, longitude and altitude, the AOI's four corners and the RVT LS's own timestamp — and emitting a POI as a second `Entity` is a modelling decision no clause of either document makes. Stated at `klv_rvt_codec.CDM_MAPPING_NOT_TAKEN` and proposed in an untracked round report, the proposal's substance being that constant. **THE WITNESS IS `carried` AND NOT `mapped`, AND IT IS THE WEAKEST IN THIS ROW SET**: ST 0806.4 prints NO worked example — Figure 7-1 is a raster image — so the ground for reading these octets is item 48's second-document agreement (ST 0601.14a §8.73 and `ST 0806.4-06` state the same sixteen octets and CRC 17945) and not a printed example. Twelve internal cross-checks stand in place of examples and establish that the four tables were transcribed consistently; they establish nothing about any value map, each of which is stated once. |
 | `74` | VMTI Local Set | None | `set` | `V` | `Event` (`DETECTION`), `Track`, `Entity` | `stanag4609 1.0.0 · mapped` | MISB ST 0903 VMTI Local Set metadata items. A nested Local Set or Pack — the items inside it are **MISB ST 0903.4**'s, park 6. **DATED NOTE, 2026-09-05: the document is now HELD and the items inside this one are transcribed and decodable** — `adapters/klv_vmti_codec.py` reads this Value into the VMTI LS, its VTargetSeries and the five nested sets under a VTarget Pack, checked against 70 of the document's own printed examples. **The row stays `not yet` anyway, and the reason is the one this table exists to make visible: nothing this adapter EMITS has moved.** No CDM object is built from item 74 by anything in this repository, so the cell that would name a mapping still has nothing to name. What is retired is the clause 'not held'; what stands is the row. See *The VMTI layer — ST 0903.4*. **PROMOTED 2026-09-05 BY THE PARK 6 ROUND, PART 2, AND IT IS THE FIRST ROW IN THIS TABLE WHOSE VALUE BECOMES AN `Event` ABOUT A THIRD PARTY.** Everything above stays in its own tense; what has changed is the sentence it ends on. Item 74's Value is now decoded by `klv_uas_codec` (§8.74's second bullet makes it a bare run of ST 0903 triplets, item 48's shape) and mapped by `adapters/stanag4609.py` under **M's ruling of 2026-09-05**. **THE WITNESS IS DOCUMENT-SIDE AND NOTHING ELSE, AND IT IS THE WIDEST ONE IN THIS TABLE**: no held stream carries an item 74 — the pinned stream's items stop at tag 65 — so what stands behind this promotion is `klv_vmti_codec.check_against_the_documents_own_examples()`, **70** of ST 0903.4's own printed examples decoded on every suite run of which **68** reproduce, the two that do not being the printed defects M ruled on. The mapping: one `DETECTION` `Event` per VTarget per packet, and a `Track` with its `Entity` **only where a VTarget carries a VTracker LS Track ID**, keyed on that UUID — "[a] value that uniquely identifies a track, using a 128-bit (16-byte) Universal Unique Identification (UUID)", §11.15.24.1, the one identifier in the document with an unconditional uniqueness statement. **The VTarget Pack Target ID Number is CARRIED IN `source_ids` AND IS NEVER A KEY**, because §11.15 resets it on tag 6, §9.4 makes tag 6 optional and `ST 0903.4-28` asks for uniqueness only "[t]o the extent possible". A target's `Position` comes from Tag 17's absolute Location pack, else from Tags 10/11's offsets added to items 23/24, else NOT AT ALL — a pixel is not a position and no frame centre means no sum, which is what Tags 10 and 11's own Notes ask for. `entity_type` is `UNKNOWN` in every case, with the VObject Ontology URI and Ontology_Class carried verbatim and resolved by nothing. **The target's CDM target cell now reads three object kinds and not `Entity.attributes`**, and the attributes half is still true — the whole decoded set rides at `attributes.vmti_local_set` on the packet `Entity` — but naming only it would understate what a consumer receives. See *The VMTI layer — ST 0903.4* and *The park 6 round, part 2* |
 | `75` | Sensor Ellipsoid Height | m | `uint16` | `2` | `Entity.position.alt_m` | `stanag4609 1.0.0` | Sensor ellipsoid height as measured from the reference WGS84 ellipsoid. SDCC-eligible: an uncertainty for this item may arrive in item 102, whose layout is ST 1010.3 and is not held — §8.75's own cell reads `Allowed in SDCC Pack? Yes`. **PROMOTED 2026-09-05 ON THE DOCUMENT-SIDE WITNESS (RULING 4), WHICH IS THE FOURTH CROSSING OF THE REOPEN CONDITION AND THE SAME ONE PARK 5 USED.** §8.75 prints its own worked example — **14,190.7195 m against the Value octets `C221`** — and `klv_uas_codec.check_against_the_documents_own_examples()` reproduces it on every suite run, decoding `C221` to 14 190.719462882427 m. The map is §8.75's own bullet, **`Map 0..(2^16)-1 to -900..19000 meters`**, read off the block and not carried: `Softval = (19900/65535) * KLVuint − 900`, an affine `uint16` and **NOT `IMAPB`**, which is why this row was not one of park 5's sixteen. One integer step is **0.30365 m**, computed from the block's own Min, Max and integer span rather than read off its `~0.3 meters` Resolution cell. **No held stream carries this tag** — the pinned stream's 26 items stop at tag 65 — so it ships against a printed example and against no held octet. **§8.75 STATES NO SPECIAL VALUES**, so unlike items 13, 14, 19 and 23 there is no reserved integer to refuse; a zero-length tag 75 is `ST 0601.14-33`'s explicit unknown and fills nothing, and a tag 75 at any length but two is a length divergence of park 13's class — this is the ONLY document-witnessed item whose block states a Required Length, which is why the codec keeps it in `AFFINE_DOCUMENT_ITEMS` and decodes it down the same path the 26 take. **PRECEDENCE AGAINST TAG 104, AND IT IS THIS REPOSITORY'S RULE STATED AS ONE.** §8.104.1's preference is for the PAIR over tag 15 — "For legacy systems, Tag 15 and Tag 75 | Tag 104 are allowed with preference for Tag 75 | Tag 104", written as a disjunction — and it states **no ordering between 75 and 104**, which was re-read off the pinned copy for this round rather than carried. So the rule is `adapters/stanag4609.HAE_PRECEDENCE`'s and says so: **104 when a packet carries both, 75 when it carries only 75, never 15.** Two grounds, both off the blocks — RANGE, because §8.104.1 states 104's purpose as "to increase the range of altitude values currently defined in Tag 75 Sensor Ellipsoid Height to support all CONOPs for airborne systems", 40 000 m against this item's 19 000 m; and RESOLUTION, 0.0078125 m at three IMAPB octets against this item's 0.30365 m. **Where both are carried and differ by more than THIS item's own LSB**, an advisory `hae_items_disagree` carries both values, their difference and the step, at `payload.klv_advisories` and at `attributes.position_basis.hae_disagreement` — the `ST 0102.10-57` / `byte_order` precedent: the selected value is still emitted, because refusing would discard a measurement over a producer's inconsistency. The threshold is one LSB and not zero because two items quantised over different ranges differ in their low bits on any real airframe. **ONE MORE FACT, READ OFF THE TWO SECTIONS TOGETHER:** §8.75 and §8.104 print the SAME KLV Key, `06.0E.2B.34.01.01.01.01.0E.01.02.01.82.47.00.00` (CRC 16670) — the extended item reuses the base item's Universal Label, so the UL does not distinguish the twins and only the tag does. **DATED NOTE, 2026-09-04 (RULING 5), NOW CLOSED.** This row read *"THIS ROW IS NOW THE ONE THING STANDING BETWEEN `alt_m` AND THE BASE HAE ITEM, AND IT WAS NOT READ THIS ROUND"* and *"blocked by nothing but scope"*. The round scoped to it read it, and both sentences are spent: nothing stands between `alt_m` and the base HAE item, and the coexistence question that note recorded for a later round is answered above. Fixtures: `tag_75_from_the_documents_own_example`, `hae_from_tag_75_when_it_is_the_only_ellipsoid_item`, `both_hae_items_agreeing_take_tag_104_and_raise_nothing`, `both_hae_items_disagreeing_raise_an_advisory_and_still_emit` and `hae_is_tag_75_and_never_tag_15s_msl` — one per clause. **MINIMUM SET, ADDED 2026-09-05 BY THE PARK 12 ROUND.** MISB ST 0902.8 Table 1 names this item, as row `15|75|104` — an ALTERNATES row, and `15`, `104` carry the same obligation. Note 1 on page 3 governs it: the members are 'governed by an “inclusive or” within MISB ST 0601' with the one exception that 'The use of Tag 75 and Tag 104 is governed by an “exclusive OR”', so the row is met when ANY member is reported and a member's own absence is not a shortfall. The obligation is §6's — 'The MISMMS is a Local Set profile of MISB ST 0601 with all items mandatory unless conditionally dependent' — discharged at the rate `ST 0902.3-04` sets, 'no less than once every thirty (30) seconds under all circumstances', with Table 2's recommended interval `Fast`. **NOT A PER-PACKET RULE**: Annex A's own Note says 'It is not mandatory that each metadata packet contain every metadata item', which is why `attributes.mismms_conformance` is an advisory and never a refusal. Transcribed at `adapters/klv_mismms.py`. |
 | `76` | Alternate Platform Ellipsoid Height | m | `uint16` | `2` | `Entity.attributes` | `not yet` | Alternate platform ellipsoid height as measured from the reference WGS84 Ellipsoid |
@@ -10490,7 +10490,8 @@ coordinates, and the RVT LS's own User Defined Time Stamp. None is mapped. Row 7
 `Entity.attributes`, and a POI is a point on the ground the aircrew nominated rather than the
 platform the packet is about; emitting it as a second `Entity`, or an AOI as a geometry, is a
 modelling decision no clause of either document makes. Stated at
-`klv_rvt_codec.CDM_MAPPING_NOT_TAKEN` and proposed in `rounds/reports/F.md`.
+`klv_rvt_codec.CDM_MAPPING_NOT_TAKEN` and proposed in an untracked round report, the proposal's
+substance being that constant.
 
 **FOUR REQUIREMENTS THIS LAYER REPORTS AND NEVER ENFORCES.** `ST 0806.4-01`..`-04` make the
 Precision Time Stamp first and the CRC-32 last in an **independent** RVT Local Set. One embedded in
@@ -10749,7 +10750,7 @@ park.
 | **4** | ~~**MISB ST 0107 — KLV Metadata in Motion Imagery**~~ **CLOSED 2026-08-26** | **0107.3** | `MISP-2015.1-08`: KLV "shall be **formatted** in accordance with MISB ST 0107". The Reason cell used to end here, distinguishing this park from park 8 on the profile's own split — ST 336 says how a triplet is *encoded* and ST 0107.3 says how KLV is formatted *in motion imagery*. **Reading the document showed that split UNDERSTATES it:** ST 0107.3 states the BER length grammar and the BER-OID chain rule, which are encoding rules by any reading. The distinction is the profile's own and is not wrong; the useful content simply was not where it put it | **CLOSED 2026-08-26.** Obtained from NSG Registry document 4738 at the pinned revision rather than the current one — the registry marks this edition superseded, and a later revision is a different document — pinned by SHA-256 `500d6752…98b69794`, 656 949 bytes, **6 pages**, and read in full. Both halves of the exit condition are discharged: the document is pinned, and the artefact it makes writable is written — for a framing standard that is a **codec**, not a row set, which is the reading the preamble above sets out. `klv_codec` now walks a UAS Datalink LS packet end to end. **What it did NOT discharge:** park 8's state, and nothing about meaning — all 141 tag rows still read `not yet`. See the framing section above and `spec/klv_pin.json`'s `length_ruling_st_0107_3` |
 | **5** | ~~**MISB ST 1201 and ST 1303 — float↔integer mapping, multi-dimensional arrays**~~ **CLOSED 2026-09-04** | **1201.3**, **1303.1**, both held | `MISP-2015.1-09` and `-10`. Every scaled numeric value inside a KLV item is mapped by ST 1201.3 and every array formatted by ST 1303.1 | Public download. The park where a wrong answer is least visible: a float mapping applied with the wrong range yields a number rather than an exception — the class of defect `cat048_codec`'s ellipsoid audit exists for. **STILL PARKED AND NOW ON ONE BLOCKER, 2026-08-27 evening.** **The acquisition half is DISCHARGED:** both documents are held and pinned at the exact revisions this cell names — **ST 1201.3**, `c5d8cb2d…bff4a07e`, 617 525 bytes, 20 pages, and **ST 1303.1**, `e30487b0…a3627895`, 1 084 929 bytes, 14 pages — obtained by the pins round after the control fetch proved the route against a recorded digest. The transient blocker was transient: three earlier attempts met **HTTP 429** at ~14:05Z and ~14:51Z, and this one met `X-RL: 0` and **HTTP 200** at 21:05Z. **THE ARTEFACT'S FIRST HALF IS WRITTEN, 2026-08-28, and this row went on denying it for a day.** `adapters/imapb_codec.py` landed in `431b0c5` — 309 lines, rows before code — under Ruling 1, and the bump gate derived **MINOR** for it with no human ruling, moving the floor to **1.3.0** by itself. The clause this replaces said the artefact was *"blocked on a per-change ruling"* and cited park 2 for *"document held, artefact unwritten"*; the plan table's own park 5 column had already recorded the landing, so the two sites disagreed. **This park is still NOT closed, and the reason is no longer a ruling.** Its artefact is a codec PLUS the tag rows that consume it, and **not one of the sixteen rows has moved** — no row of this file cites the codec in the name-and-version form every shipped adapter's rows carry, and this sentence deliberately does **not** spell that form, because spelling it makes this cell the file's only occurrence and disarms the very `grep` the claim invites (sweep rule 9, `synapse_cdm/README.md`; found by rule 11's first live sweep, which is how a claim that read as checkable turned out to have been checking itself) — because none of the sixteen is witnessed by the pinned stream and the scope contract blocks them independently of this park. So the state is a NEW one and is not park 2's precedent: document held, half the artefact written, and the remainder blocked by a different rule. **DATED NOTE, 2026-09-04: park 2 CLOSED**, so the precedent this cell distinguishes itself FROM is now a discharged one — which sharpens the distinction rather than dissolving it. Park 2 closed by writing the row set its document supports; this park's remaining half is sixteen rows the scope contract blocks independently, and no closure of park 2 reaches them. Repaired 2026-08-29 by the disjunction sweep. **Narrowed from held documents instead.** ST 0601.14a's reference list [13] pins it as "MISB ST 1201.3 Floating Point to Integer Mapping, **Oct 2017**", corroborating the version cell from the governing document rather than from the delegation table alone. **Its reach is 14 rows by the Format column and 16 by the sections, and the two numbers are both right — which is why both are stated.** Fourteen of the 141 Table 1 rows have a Format column reading `IMAPB`: tags 96, 103, 104, 105, 109, 112, 113, 114, 117, 118, 119, 120, 132 and 134. **The sixteen this file's ledger entry enumerates are those fourteen plus 128 and 130**, whose Format column reads `vlp` and whose §8.x sections map pack MEMBERS with IMAPB — §8.128 states "Wavelength Min and Wavelength Max values are IMAPB(0,1e9,4)" and §8.130 "Each location is described in a DLP containing IMAPB". **So IMAPB reaches inside packs, and the Format column does not show it.** The ledger counts sections that name IMAPB and this cell counts rows whose Format IS IMAPB; the sets differ by exactly those two tags, and the larger number is the honest measure of the park. Re-derived 2026-08-27 from ST 0601.14a's own table and sections. **Not one of the sixteen is among the 26 items the pinned stream witnesses**, whose formats are `uint8/16/32/64`, `int16/32` and `utf8` throughout and whose highest tag is 65. So this park blocks rows and no held octet. **ST 1303 is a weaker dependency than its row implies:** `1303` occurs **zero** times in ST 0601.14a's 218 pages, so the governing document delegates nothing to it and the delegation is MISP-2019.1's alone. **CLOSED 2026-09-04 BY THE PARK 5 ROUND, ON BOTH HALVES OF `shared_exit_condition` FOR ST 1201.3 AND ON A DISCHARGE-BY-ABSENCE FOR ST 1303.1 — AND THE TWO HALVES OF THIS PARK'S TITLE CLOSED FOR DIFFERENT REASONS, WHICH IS WHY BOTH ARE STATED.** **ST 1201.3, FIRST HALF** — discharged 2026-08-27 and unchanged: obtained at the exact revision this row's Version cell names, pinned by SHA-256 `c5d8cb2d…bff4a07e`, 617 525 bytes, 20 pages, title-page identity read; re-digested from disk at the start of the closing round and matching. **ST 1201.3, SECOND HALF — the artefact, and it took two rounds because the artefact is a codec PLUS the rows that consume it.** `adapters/imapb_codec.py` landed 2026-08-28 in `431b0c5` and the sixteen rows stayed `not yet` for a week, which is the state this cell recorded above as *"document held, half the artefact written, and the remainder blocked by a different rule"*. **RULING 1 (2026-09-04) established that the different rule was already satisfied.** The scope contract's own reopen condition — *"a second pinned stream, OR a document-side check as strong as a worked example — and ST 0601.14a prints one per item"* — was written into the "Not witnessed" ledger row on 2026-08-26, and the codec had been reproducing all fourteen printed worked examples on every suite run since it landed. **FIFTEEN of the sixteen rows now read `stanag4609 1.0.0`** — the fourteen IMAPB items plus tag 128, whose pack layer is `adapters/klv_pack_codec.py` — and this cell's own outstanding clause is the clause discharged: it read *"not one of the sixteen rows has moved — no row of this file cites the codec in the name-and-version form every shipped adapter's rows carry"*, and fifteen of them now do. **The sentence that deliberately refused to spell that form is retired with the claim it protected**, and the `grep` it was protecting is now pointless for the opposite reason: the form occurs fifteen times in this file's ST 0601 row set. **THE SIXTEENTH ROW DID NOT MOVE AND THE PARK CLOSED ANYWAY, WHICH IS A DISTINCTION THIS TABLE HAS NOT HAD TO DRAW BEFORE.** Tag 130 Airbase Locations stays `not yet` — §8.130 prints no worked example, so RULING 1's condition is unmet for it, and its HAE member's range is stated twice in one section 300 m apart. **That row is not blocked by ST 1201.3.** The codec maps every member the section names; what is missing is a statement of one member's `a` and a printed example to check it against, both of which are ST 0601's to supply. So the row is carried under the scope contract in the ledger and this park does not stay open for it — the exit condition is the document plus the artefact the document makes writable, and a row the artefact can already map is not a missing artefact. **ST 1303.1 — DISCHARGED BY ABSENCE (RULING 2), WHICH IS THE PARK 4 READING APPLIED TO A DOCUMENT THAT MAKES NOTHING WRITABLE HERE.** Park 4 read the third step as *the artefact this document makes writable* — a row set for a field dictionary, a codec for a framing standard. ST 1303.1 is held and pinned (`e30487b0…a3627895`, 1 084 929 bytes, 14 pages, re-digested clean in the closing round). **Re-derived from the pinned bytes on 2026-09-04 under `gates/pdf_text.py`'s join-and-collapse rule: `1303` occurs ZERO times in ST 0601.14a's 218 pages, and so does `MDARRAY` — zero for the spaced and lower-case variants too.** So no row of any held row set consumes a multi-dimensional array, the row set this document supports is **within this repository's scope empty**, and there is no artefact for it to make writable. **That is a discharge and not a silent drop, and it carries a reopen condition: a held row set that names ST 1303.** Park 6, MISB ST 0903.4's VMTI and Track Metadata, is the candidate — it is the row set most likely to carry an array — and if it lands naming ST 1303, this half reopens on its own terms with the document already held. **WHAT THIS CLOSURE DOES NOT DISCHARGE.** **Not one held octet is newly understood** — all fifteen promoted rows ship against ST 0601.14a's printed worked examples and against no wire anybody here has met, and each row says so in its own description. **Not the first half of the reopen condition**: a second pinned stream has never been obtained, and it stays the condition for the other 99 rows. **Not tag 75**, the base HAE item, which is `uint16` and never was one of the sixteen — RULING 5 records what §8.104.1 says about the two coexisting and leaves it for a round scoped to it. **Not park 3's epoch, not park 6, 7, 10 or 12's documents, and not park 11's artefact.** **And not the printed `Resolution` cells of §8.104, §8.105 and §8.112**, which the closing round proved wrong from those sections' own example octets — a factor of ten twice and a digit slip once; the codec computes the step from `(a, b, L)` and reads no such cell, so the defect is recorded at `imapb_codec.PRINTED_RESOLUTION_DISAGREEMENTS` rather than worked around |
 | **6** | ~~**MISB ST 0903 — VMTI and Track Metadata**~~ **CLOSED 2026-09-05** | **0903.4** | §4.4.2.4. The detections and derived tracks live here, which is the content the CDM is shaped for | Public download. **STILL PARKED AND STILL THE HIGHEST-VALUE PARK AFTER 1**, and this cell is a REPAIR of its second half only. It used to read *"what would make this adapter emit `Event`s of type `DETECTION` rather than only parked bytes"* — a comparison written against an adapter that had not shipped yet. **The `DETECTION` half STANDS and is measured, not assumed:** running adapter #10 over the pinned stream this round emits `STATUS_CHANGE` on all six events and `DETECTION` on none, and item 2's sibling rows say why — nothing in the 26 witnessed items detects anything, and the item that would is tag 74, this park's own. **"Only parked bytes" DOES NOT STAND.** Re-derived from the adapter this round rather than carried: the six packets yield **six `Entity`/`Event` pairs, twelve objects**, and every one of the six entities is a `PLATFORM` carrying a **GNSS `Position`** — 54.681323°/−110.168560° at the first packet, 54.652356°/−110.108494° at the sixth — and a `Kinematics` whose `speed_mps` is filled on all six at **46.0, 44.0, 43.0, 43.0, 43.0 and 47.0**. **The six values are the repair's own point:** the last sweep recorded this refutation with the single figure `speed_mps=46.0`, which is packet 0's alone, and a park priced on "only parked bytes" is not corrected by one example any more than it is established by none. **What is still true is narrower and is worth more than the sentence it replaces:** this adapter emits a platform reporting itself, six times, and no observation OF anything — so what park 6 buys is not the adapter's first canonical field but its first `Event` about a third party. `course_deg` is `None` on all six and `climb_mps` on all six; those are park 5's and the scope contract's, not this park's. Repaired by the decay sweep of 2026-08-28 **STILL PARKED, AND NARROWED TO ONE HALF OF ONE THING, 2026-09-05 BY THE PARK 6 ROUND, PART 1.** The document is HELD: MISB ST 0903.4, 23 October 2014, `da46ba5e…c57e442b`, 2 952 504 bytes, 113 pages, pinned at `delegated_specifications_held.st_0903_4` after both official routes were asked first and refused, and this row's `delegation_table` entry moved from `held: false` to `held: true` the same day. **The first half of the artefact is written**: the fourteen element tables are transcribed with four cross-checks — 148 of 148 printed CRCs recomputed from their own key bytes, 0 Appendix-A ULs absent from the tables, 0 CRC disagreements, 97 of 97 requirement IDs distinct and gapless — and `adapters/klv_vmti_codec.py` reads item 74's whole hierarchy, checked against **70** of the document's own printed examples of which **68** reproduce. **WHAT KEEPS THIS PARK OPEN IS THE MAPPING AND NOTHING ELSE.** The exit condition is the document plus the artefact the document makes writable, and here the artefact has two halves: the codec, which is written, and what a VTarget BECOMES in the CDM, which is not. That second half is a decision about what a consumer receives — a `DETECTION` `Event` per target per packet, an `Entity` per target identity, a `Track` from an id across packets, or some of each — and the document narrows every candidate rather than settling it: the Target ID Number persists *"until the identification number is reset by the New Detection Flag (Tag 6)"*, VObject's type fields are a URI and a free string in a register this repository does not hold, and §10 **discourages the only track carrier item 74 can reach** in favour of the VTrack LS, which item 74 can never carry. So the round wrote the fork up rather than settling it, and **row 74 stays `not yet`** because nothing this adapter emits has moved. **WHAT THIS ROUND ALSO SETTLED FOR ANOTHER PARK:** `ST 1303` occurs ZERO times in these 113 pages and `MDARRAY` zero, so park 5's `st_1303_1_reopen_condition` — which named this park as its candidate — is NOT met, and the discharge-by-absence is strengthened by the one row set that could have reopened it. **AND THREE PRINTED DEFECTS ARE RECORDED RATHER THAN WORKED AROUND**, under M's ruling that a table is normative and a worked example an illustration: the VMask Bit Mask octets (each 256 high against the section's own 16×9 figure), the VTracker Track ID's `[K]`/`[L]` framing (reversed against its own row's Tag and Length cells), and five Appendix A entries labelling their owner `VTrack Pack`, a construct this document does not define. See *The VMTI layer — ST 0903.4* **CLOSED 2026-09-05 BY THE PARK 6 ROUND, PART 2, AND WHAT CLOSED IT IS THE ONE HALF THE CELL ABOVE NAMES.** That cell says *"what keeps this park open is the mapping and nothing else"* and lists the four candidates the document narrows without settling. The mapping is now written and it is **M's ruling of 2026-09-05 rather than a reading**, which is the honest description: `adapters/stanag4609.py` emits one `DETECTION` `Event` per VTarget per packet, and a `Track` with its `Entity` **only where a VTarget carries a VTracker LS Track ID**, keyed on that UUID and on nothing else. **THE IDENTIFIER THE OBVIOUS MAPPING WOULD HAVE USED IS REFUSED, AND THREE CLAUSES REFUSE IT**: §11.15 scopes the Target ID Number *"until the identification number is reset by the New Detection Flag (Tag 6 within the VTarget Pack)"*; §9.4 with `ST 0903.4-09`/`-10` makes every TLV triplet including tag 6 optional, so the reset need not be observable at all; and `ST 0903.4-28` requires uniqueness only *"[t]o the extent possible"*, with §11.15 adding that systems *"may use the same Target ID Number to identify a common target detected by different sensors"*. The number is carried in `source_ids` under its own system name and is never a key. **THE CARRIER THAT IS USED IS ONE §10 DISCOURAGES AND THAT IS RECORDED RATHER THAN ARGUED AWAY** — the recommended VTrack LS is outside item 74's reach by §9.1, so the discouraged carrier is the only one there is. **WHAT THE PARK BOUGHT, MEASURED**: five fixtures, 106 harness verdicts still 0 failed, and the adapter's first `Event` about a third party — `DETECTION` ×0 before this round and ×1, ×1, ×1, ×1, ×2 across the five new payloads. **AND THE PINNED STREAM DID NOT MOVE**: it carries no item 74, this round declared no per-object annotation for a packet without one, and all 96 goldens written before it are byte-identical after it — which is the strongest form of the brief's condition rather than the permitted one. **WHAT THIS PARK DID NOT CLOSE**: no held stream carries an item 74, so every octet the mapping reads is a printed example's, and Table 16's merge and split states are carried as the producer's statement with no consequence drawn — a one-UUID-one-`Track` mapping cannot express either, and saying so is the honest end of the round rather than a fifth candidate |
-| **7** | ~~**MISB ST 0806 — Remote Video Terminal Metadata Set**~~ **CLOSED 2026-09-06** | **0806.4**, held | §4.4.2.4: *"MISB ST 0806 [59] defines KLV Metadata that supports a Remote Video Terminal (RVT)"* | Public download. **NARROWED 2026-09-06 BY THE PARK 7 ROUND AND STILL OPEN, IN PARK 6's RECORDED SHAPE: the document is held, the transcription is written and the codec is written and WIRED, and what is left is this row's own strikethrough.** The document is pinned at `klv_pin.json`'s `delegated_specifications_held.st_0806_4` (sha256 `58de4265…d6e01037`, 716 099 bytes, 13 pages, cover *'MISB ST 0806.4'* / *'27 February 2014'*); the four tables are transcribed in this document — Table 8-1's RVT LS at 21 tags, Table 8-2's POI LS at 10, Table 8-3's AOI LS at 10, Table 8-4's User Defined LS at 2; `adapters/klv_rvt_codec.py` walks all four with ONE recursive `decode_set`; `klv_uas_codec` delegates item 73 to it and `adapters/stanag4609.py` carries the decoded set into `Entity.attributes.rvt_local_set`. **ROW 73 IS PROMOTED, WHICH IS WHERE THIS DIFFERS FROM PARK 6's NARROWING** — park 6's round left the codec wired into nothing and row 74 at `not yet`; this one emits, so the row says so. **WHY THE STRIKETHROUGH IS NOT HERE, AND IT IS NOT A DOUBT ABOUT THE DOCUMENT.** Closing this row takes the table to ONE open row, and at one open row FIVE GUARDS STOP BEING EXERCISABLE AGAINST THE REAL TABLE: `gates/parks_table.py --mutation-check` case 1 and `test_the_set_claim_guard_is_not_vacuous_in_either_direction` both need a set-claim naming an OPEN park, and no honest claim can name one — a multi-member claim would have to pair row 10 with a closed row and would fail CLOSED MEMBER on its face; `check_stated`'s MISSING branch cannot fire when one open row is claimed by both halves of a partition; `test_self_membership_is_an_observation_and_never_a_problem` and `test_an_empty_spec_directory_reports_unverifiable_and_never_absence` fail for the same emptiness, the latter because the last open row's blocker is the one this table records as NOT DERIVABLE from a filename. **THE OBVIOUS REPAIR IS REFUSED BY A DOCUMENTED DECISION**: exercising case 1 against a synthetic claim is what the vacuity test's own docstring rules out — *"Mutating the real table rather than a synthetic one, because a fixture that spells the shape is a fixture that can drift away from it"*, citing `synapse_cdm/README.md`'s sweep rule 1. So how these five guards should behave on a one-row table is a decision about what they MEAN, it is a fork this round's brief did not pre-rule, and it is written up for M in `rounds/reports/F.md` rather than guessed at here. **DATED CORRECTION 2026-09-06: THE PARAGRAPH ABOVE IS LEFT IN ITS OWN TENSE AND THIS ROW IS NOW CLOSED.** M ruled the fork the same day, and the ruling is not the repair the paragraph refused: a guard whose shape this table no longer carries **reverses, in its own copy only, the most recent recorded closure that restores that shape** — a real row, its real claim, its real blocker — and plants its mutation on that, exactly as before. Which closure is reversed is derived from this table at run time and never typed in a guard, which is the drifting fixture one level up; the claim put back beside the reopened row is one THIS TABLE WITHDREW, read back through the `rows`-for-`parks` convention this file has used since 2026-09-04; and the guard's message names the closure it reversed, because a check that quietly re-shapes its subject is a check whose green means something else. It is `gates/parks_table.py`'s `reversed_to`, asked for by shape through `SHAPES`, and it holds at ZERO open rows as well as at one — the state this table is built to reach — which `test_the_reversal_still_restores_every_shape_when_no_row_is_open_at_all` demonstrates on a copy. **Closed on the document and the artefact both**, as this row's own condition asked: ST 0806.4 is pinned, its four tables are transcribed, `adapters/klv_rvt_codec.py` reads them and row 73 is promoted. **REOPEN CONDITION**: a later edition of ST 0806 delegated by a profile this record holds (the delegation table pins .4 and MISB has published .5), or an element of any of the four sets this codec refuses rather than carries, or the CDM mapping proposal in `rounds/reports/F.md` being ruled — the elements ride in `Entity.attributes` and no `Entity` and no geometry are derived from them, which is a decision recorded at `klv_rvt_codec.CDM_MAPPING_NOT_TAKEN` and not a shortfall of the document |
+| **7** | ~~**MISB ST 0806 — Remote Video Terminal Metadata Set**~~ **CLOSED 2026-09-06** | **0806.4**, held | §4.4.2.4: *"MISB ST 0806 [59] defines KLV Metadata that supports a Remote Video Terminal (RVT)"* | Public download. **NARROWED 2026-09-06 BY THE PARK 7 ROUND AND STILL OPEN, IN PARK 6's RECORDED SHAPE: the document is held, the transcription is written and the codec is written and WIRED, and what is left is this row's own strikethrough.** The document is pinned at `klv_pin.json`'s `delegated_specifications_held.st_0806_4` (sha256 `58de4265…d6e01037`, 716 099 bytes, 13 pages, cover *'MISB ST 0806.4'* / *'27 February 2014'*); the four tables are transcribed in this document — Table 8-1's RVT LS at 21 tags, Table 8-2's POI LS at 10, Table 8-3's AOI LS at 10, Table 8-4's User Defined LS at 2; `adapters/klv_rvt_codec.py` walks all four with ONE recursive `decode_set`; `klv_uas_codec` delegates item 73 to it and `adapters/stanag4609.py` carries the decoded set into `Entity.attributes.rvt_local_set`. **ROW 73 IS PROMOTED, WHICH IS WHERE THIS DIFFERS FROM PARK 6's NARROWING** — park 6's round left the codec wired into nothing and row 74 at `not yet`; this one emits, so the row says so. **WHY THE STRIKETHROUGH IS NOT HERE, AND IT IS NOT A DOUBT ABOUT THE DOCUMENT.** Closing this row takes the table to ONE open row, and at one open row FIVE GUARDS STOP BEING EXERCISABLE AGAINST THE REAL TABLE: `gates/parks_table.py --mutation-check` case 1 and `test_the_set_claim_guard_is_not_vacuous_in_either_direction` both need a set-claim naming an OPEN park, and no honest claim can name one — a multi-member claim would have to pair row 10 with a closed row and would fail CLOSED MEMBER on its face; `check_stated`'s MISSING branch cannot fire when one open row is claimed by both halves of a partition; `test_self_membership_is_an_observation_and_never_a_problem` and `test_an_empty_spec_directory_reports_unverifiable_and_never_absence` fail for the same emptiness, the latter because the last open row's blocker is the one this table records as NOT DERIVABLE from a filename. **THE OBVIOUS REPAIR IS REFUSED BY A DOCUMENTED DECISION**: exercising case 1 against a synthetic claim is what the vacuity test's own docstring rules out — *"Mutating the real table rather than a synthetic one, because a fixture that spells the shape is a fixture that can drift away from it"*, citing `synapse_cdm/README.md`'s sweep rule 1. So how these five guards should behave on a one-row table is a decision about what they MEAN, it is a fork this round's brief did not pre-rule, and it is written up for M in `rounds/reports/F.md` (an untracked round report, not in this repository) rather than guessed at here. **DATED CORRECTION 2026-09-06: THE PARAGRAPH ABOVE IS LEFT IN ITS OWN TENSE AND THIS ROW IS NOW CLOSED.** M ruled the fork the same day, and the ruling is not the repair the paragraph refused: a guard whose shape this table no longer carries **reverses, in its own copy only, the most recent recorded closure that restores that shape** — a real row, its real claim, its real blocker — and plants its mutation on that, exactly as before. Which closure is reversed is derived from this table at run time and never typed in a guard, which is the drifting fixture one level up; the claim put back beside the reopened row is one THIS TABLE WITHDREW, read back through the `rows`-for-`parks` convention this file has used since 2026-09-04; and the guard's message names the closure it reversed, because a check that quietly re-shapes its subject is a check whose green means something else. It is `gates/parks_table.py`'s `reversed_to`, asked for by shape through `SHAPES`, and it holds at ZERO open rows as well as at one — the state this table is built to reach — which `test_the_reversal_still_restores_every_shape_when_no_row_is_open_at_all` demonstrates on a copy. **Closed on the document and the artefact both**, as this row's own condition asked: ST 0806.4 is pinned, its four tables are transcribed, `adapters/klv_rvt_codec.py` reads them and row 73 is promoted. **REOPEN CONDITION**: a later edition of ST 0806 delegated by a profile this record holds (the delegation table pins .4 and MISB has published .5), or an element of any of the four sets this codec refuses rather than carries, or the CDM mapping proposal (in an untracked round report, the constant named next carrying its substance) being ruled — the elements ride in `Entity.attributes` and no `Entity` and no geometry are derived from them, which is a decision recorded at `klv_rvt_codec.CDM_MAPPING_NOT_TAKEN` and not a shortfall of the document |
 | **8** | ~~**SMPTE ST 336 — Data Encoding Protocol Using Key-Length-Value**~~ **CLOSED 2026-09-03** | **ST 336:2017**, held | `MISP-2015.1-07`. The KLV encoding itself — key form, length form, the triplet. **CLOSED 2026-09-03 by the publisher round, and it is the only park in this table that was closed by acquiring a document the row itself had priced as unobtainable without a budget.** **Both halves of the shared exit condition, and it needed both.** **FIRST HALF:** the exact edition this row's Version cell names — ST 336:2017, which is what MISP-2019.1's Appendix B reference [13] pins — obtained from SMPTE's own document library at `pub.smpte.org/doc/st336/`, unpacked outside the tree from the publisher's zip, and pinned by SHA-256 `d6658f63…f37dc294`, **457 808 bytes**, **36 pages**, title page read. **ST 336:2007 was acquired in the same pass and pinned beside it** — SHA-256 `e05e5415…4ea92522`, **300 320 bytes**, **41 pages** — because register entry **KLV 11** is a question about two editions and could not be answered from one. It is **not** this park's document and not a delegation the profile makes: three held documents require conformance to it, which is a status of its own. **SECOND HALF:** the artefact the document makes writable, which for a framing standard is a **codec** and not a row set — park 4's reading of that step, applied a second time. §5.3 states what `0x80` means, so `klv_codec.decode_ber_length` now refuses it as a **framing error citing ST 336** where it used to park it as underivable, with a refusal fixture behind the change. **WHAT THE ROW USED TO SAY AND WHY IT IS GONE.** This cell called ST 336 paywalled and named this park the one that needs a budget. That premise was refuted on 2026-09-03 by the maintenance round, which asked the publisher's own host and got a 200 — and this round paid nothing, because there was nothing to pay. **The profile's FORWARD sentence about commercial references is a quotation and stays one**; what moved is the inference this repository drew from it and never asked the publisher to confirm. **The policy behind the 200, recorded with the instant it was read (rule 12):** SMPTE announced in June 2026 that its published Standards, RPs, EGs and RDDs are freely accessible at `pub.smpte.org/doc/`. `www.smpte.org/setting-the-standards-free` answered **200** at **2026-09-03T21:25:43Z**, headlined *"SMPTE Makes Its Standards Freely Accessible, Opening Standards Library to the Global Media Technology Community"* and linking `pub.smpte.org/doc/`. **Free access is not redistribution and nothing here redistributes**: `.gitignore` carries `*.pdf` and `*.zip`, both new files were checked with `git check-ignore -v` and report `.gitignore:42:*.pdf`, and `tests/test_cdm_pins.py` asserts no PDF is tracked anywhere | **CLOSED 2026-09-03.** **What the two residual absences turned out to be, which is what actually closed this park — the download only made the reading possible.** This park's residue was two absences in ST 0107.3, and ST 336:2017 answers them **differently from each other**. **(1) `0x80` as a first length octet — STATED, at §5.3**, normatively and in the body rather than in a note: *"In such cases, the Length field shall be set to [0x80] which shall indicate a non-deterministic length of the Value field. Any application document which allows the length of the Value field to be undefined shall define an alternative method of locating the end of the Value field."* NOTE 3 gives the reason for the choice — the value *"is normally meaningless as an ASN.1 BER long-form value as it indicates zero subsequent bytes"*. **No held MISB document defines such an alternative method**, so a MISB local set carrying `0x80` has no conforming way to be terminated and the decoder now says so. **(2) Any ceiling on the count of length octets — DELEGATED ONWARD, and ST 336 says so in its own voice.** §5.3's NOTE 1 reads *"While there are no restrictions in this standard on the maximum number of bytes in the Length field…"*, and the normative sentence sends the length rules to **ISO/IEC 8825-1 §8.1.3, 8.1.3.3 through 8.1.3.5** — X.690, named with paragraph numbers in the Normative References. So the absence was never ST 336's to fill: it imposes no ceiling, it says it imposes none, and it points at the document that does. **That is where register entry KLV 13 lands and closes** — its reasoning was right, its pointer at X.680 was the error, and ST 336:2017's own reference list adjudicates it. **WHAT THIS CLOSURE DOES NOT DISCHARGE.** Not one tag row: all 141 still read `not yet`, and none was eligible — a framing rule says where an item begins and never what it means. Not park 3's epoch, not park 5's `IMAPB` ranges, not park 2's unwritten row set, and no other park's document. **Not the X.690 ceiling itself:** ST 336:2017's Annex I reproduces §8.1.3.5 verbatim and that text forbids an initial octet of `0xFF`, which would cap the count at **126** rather than the structural 127 the codec uses — **recorded and deliberately not enforced**, because X.690 is not held and an informative annex quoting an unheld standard is not this repository's authority for a refusal. **And not the guard's tier.** The suite check on this row asserted a fact that lives at `pub.smpte.org` and could not reach it; it now asserts a closure whose evidence is two pinned PDFs on disk. That is a real improvement and a small one: it was **protocol-gated** on the day it was refuted and is **document-gated** now, and in neither state was it ever the *suite-gated* check its label claimed |
 | **9** | ~~**MISB ST 1402 — MPEG-2 Transport Stream for Class 1/Class 2**~~ **CLOSED 2026-08-27** | **1402.2**, obtained and pinned | `MISP-2015.1-48` and `-49`, §3.6.9.1 and §5.3.1. Names how KLV is carried in the multiplex, and the two multiplex methods | **CLOSED 2026-08-27, and it is the first park closed by an ACQUISITION that earlier rounds could not make.** Both halves of the exit condition are discharged: the document is pinned above — SHA-256 `52a3b32a…773100e0`, 1 112 404 bytes, 13 pages — and the artefact half, which for this park was prose and nothing else, is the 26-requirement row set below. **What actually moved was the HOUR, not the record's understanding.** Three prior attempts asked at ~14:05Z and ~14:50Z and met HTTP 429; this one asked at 20:20Z and met `X-RL: 0` and HTTP 200. The quota is evidently windowed in a way no earlier round had varied, because every earlier attempt held the hour fixed while varying nothing else. **The pin was TESTED before anything new was trusted:** the byte-exact archived ST 0601.4 URL was re-fetched in full first and digested to this record's pinned value exactly, so the route was shown to serve the recorded bytes before it was asked for unrecorded ones. **The two official routes were asked first and neither served it** — `gwg.nga.mil` HTTP 403 with an S3 `AccessDenied` body, `nsgreg.nga.mil` HTTP 200 whose body is the F5 JavaScript interstitial, which is a challenge page and not a document — so the provenance is the Internet Archive's snapshot of `gwg.nga.mil`'s own copy, the same 2020-01-11 crawl that served EG 0601.1 and ST 0601.4, and the caveat those pins carry governs this one. **What it did NOT close:** no other park, and no tag row — all 141 still read `not yet`. Settlement 1's asymmetry is unchanged; what this buys is that widening the input to a transport stream is no longer blocked on an unobtained document |
 | **10** | ~~**MISP-2019.1: Motion Imagery Handbook**~~ **CLOSED 2026-09-06** | the 2019.1 edition, Nov 2018 | STANAG 4609 Ed 5 lists it as its only OTHER RELATED DOCUMENT and MISP §1.3 calls it a companion "providing definitions of terms used with more background and technical detail". But §3.6.7 says of the same document that it "defines the Structure of the Common Metadata System (CMS), describes how to organize the sensor/platform data into a hierarchy of KLV Packs and Local Sets ... **and defines the required data items**" — register entry **KLV 8**. **DATED ADDITION 2026-09-06: THE DOCUMENT IS HELD AND KLV 8 IS RULED. The sentences above are left in their own tense and the question they pose is ANSWERED — against the reading they lean toward.** The Handbook was acquired by hand and pinned (SHA-256 `6754f528…c6531369`, 4 356 271 bytes, **124 pages**), and §3.6.7's sentence was tested against it limb by limb. **It holds on one limb of three.** The Handbook does describe a hierarchy of KLV Packs and Local Sets — §7.5 defines *"KLV Root: The top-most parent of a KLV hierarchy"* — but it uses the term **Common Metadata System nowhere** (zero occurrences, as is "CMS"), and it **defines no required data items** (zero occurrences of the phrase; all eight occurrences of "required" read individually, none about a data item; **one** "shall" in 124 pages, and it is *"The term FMV shall not be used in contractual language"*). **The sentence that decides is the Handbook's own Scope, PDF page 10:** *"The MISP succinctly states requirements, while the Motion Imagery Handbook discusses principles underlying requirements more thoroughly."* **So it is a companion — normative for TERMS and for nothing else, and it says that too**: *"there is no single authoritative source for technical definitions of terms within the community; therefore, the Motion Imagery Handbook serves as the authoritative source of definitions for the MISB community of practice."* Counts taken over text joined and whitespace-collapsed under `gates/pdf_text.py`'s rule, with the reader validated against this record's own recorded page counts and against §3.6.7 read off the pinned profile. | Public download, the same two URLs. Its billing and its stated function disagree, so whoever obtains it answers that first: a document that defines required data items is normative in fact, whatever the wrapper's headings say  **DATED ADDITION 2026-09-06: THE PLAN IS DISCHARGED AND THIS ROW IS STILL OPEN, FOR A REASON THAT IS NOT ABOUT THIS DOCUMENT.** Both public URLs were asked again and both refused (403 with an S3 body; the `www` alias no longer resolves), the registry served the F5 interstitial at HTTP 200 on two routes, and the archive holds no capture of any 2019.x Handbook — so the acquisition was **BY HAND**, which is the gate the plan names for exactly this case. What is left is the strikethrough, and it is **held back by a guard, not by a question**: closing this row takes the table to **zero open rows**, where `tests/test_cdm_parks_table.py::test_the_table_is_found_and_has_the_shape_a_row_check_needs` asserts *"a table with no open parks makes the blocker check vacuous"*. That assertion has to be **INVERTED rather than reversed** — `gates/parks_table.py`'s `reversed_to` answers the four mutation shapes at zero open rows, and this is the fifth shape, which it does not answer — and inverting it changes what a guard MEANS. **No ruling for that exists, so this round wrote the reading and left the strikethrough**, on the same discipline the park 7 round used when it declined to close on a synthetic claim. **Reopen/close condition: a ruling on that one assertion.** Everything else this row ever asked for is done. **DATED ADDITION 2026-09-06, LATER THE SAME DAY: THE RULING CAME AND THIS ROW IS STRUCK THROUGH.** The guard was the whole of what was left, and it is ruled the way the park 7 round's own ruling already pointed — the guards must hold at zero open rows too, since this table ends all-closed. `test_the_table_is_found_and_has_the_shape_a_row_check_needs` no longer asserts that an open row exists; it asserts the two things that assertion was for — the table is well-formed, thirteen contiguous rows, and `gates/parks_table.py`'s reversal machinery still reaches this table's shape, which it does through a fifth `SHAPES` entry, *"an open row exists at all"*, answered by reversing THIS closure in the guard's own copy and naming it in the message. The other reading was refused by name: a park that can never be struck through by construction is a defect in the guard and not a property of the table. **The gate reads 13 rows, 0 open, 13 closed, 0 set-claims, 0 failed**, and `--mutation-check` no survivors. Nothing about the document changed between the reading above and this line — what changed is that the record can now say the delegation table is fully answered |