feat: Link-Taxonomie abgeschlossen - Lint hart ab kb_version 4.0.0, outbound: an Type-Spec gebunden, part-of/composition als Inversenpaar (4.1.0, #40)

Files changed:
- CHANGES.md
- VERSION
- instructions/link-taxonomy.md
- instructions/migrations/4.0.0-link-taxonomy.md
- kb/comparisons/amd-pstate vs acpi-cpufreq.md
- kb/concepts/Episodic Memory.md
- kb/concepts/Memory Lifecycle.md
- kb/concepts/Mesh Sync.md
- kb/concepts/Procedural Memory.md
- kb/concepts/Reciprocal Rank Fusion.md
- kb/concepts/Semantic Memory.md
- kb/concepts/Shared vs Private.md
- kb/concepts/Split Threshold.md
- kb/concepts/Stub Threshold.md
- kb/concepts/Supersession.md
- kb/concepts/Typed Relationships.md
- kb/concepts/Vector Search.md
- kb/concepts/Work Coordination.md
- kb/concepts/Working Memory.md
- kb/entities/technologies/Wine-Staging.md
- kb/entities/tools/pascalandy schema.md
- kb/index.md
- kb/log.md
- kb/sources/COLLECTION.md
- tools/CONTRACT.md
- tools/chemenu/commands/lint.py
- tools/chemenu/kb_collections.py
- tools/chemenu/lint_core.py
- tools/chemenu/tests/test_conventions.py
- tools/chemenu/tests/test_lint.py
- tools/chemenu/tests/test_type_resolver.py
- types/comparison.md
- types/comparison.schema.yaml

instructions/link-taxonomy.md

@@ -40,9 +40,18 @@ edge merely to mirror the first one.** The inbound view is rendered from the gra
 `index rebuild` and `search`, so a reader landing on the target sees what points at it whether
 or not anyone wrote a second edge.
 
-That is why most labels below have no inverse. Only two pairs do, because in each the reverse
+That is why most labels below have no inverse. Only three pairs do, because in each the reverse
 direction is a genuine primary statement someone would write on its own: `depends-on` /
-`required-by` and `runs-on` / `hosts`.
+`required-by`, `runs-on` / `hosts`, and `composition` / `part-of`.
+
+The third was added after the 4.0.0 migration, from measurement rather than from the desk. A
+parent-child structure - a tier list and its tiers, a spectrum and its levels - produces the
+question on nearly every page: the parent writes `composition`, and the child then reaches for
+either `part-of` or `see-also`. The migration run answered `see-also`, on the reading that
+`part-of` would be a mirror, and left sixteen edges saying "these two are related" about a
+relationship the catalogue already had a word for. It is not a mirror: the parent's sentence
+lists its parts, the child's names the whole it belongs to, and a reader landing on the child
+needs the second one.
 
 ## When to run
 
@@ -129,6 +138,13 @@ Inference and comparison between ideas.
 | `composition` | is composed of the target |
 | `part-of` | is a component of the target |
 
+`composition` / `part-of` is the third **inverse pair**, alongside `depends-on` / `required-by`
+and `runs-on` / `hosts` in the operational register. Being a pair does not make the second edge
+obligatory - direction is still authored - it settles *which label* the second edge takes when
+someone does write it. The child of a `composition` writes `part-of`, not `see-also`: what it
+is a component of is a primary statement about the child, and `see-also` says strictly less
+about the same fact.
+
 `grounds` / `rests-on` is a genuine pair and both directions are primary statements; they are
 listed separately rather than as inverses because either page may legitimately carry only its
 own side.

instructions/migrations/4.0.0-link-taxonomy.md

@@ -36,8 +36,9 @@ you know the run is finished.
 **Nothing breaks while it is outstanding.** Unlabelled edges and undelimited regions are read,
 not rejected: `links.py` treats a bare title as an edge whose label is not declared yet, and
 `provenance.split_cite_block` falls back to the pre-marker layout. That is deliberate - a corpus
-has to stay readable while it is being converted - and it is why the two lint findings are
-advisory until step 6 promotes them.
+has to stay readable while it is being converted - and it is why the two lint findings stay
+advisory for as long as `kb_version` is below 4.0.0, which is exactly as long as this document
+is outstanding.
 
 ## Steps
 
@@ -102,7 +103,7 @@ advisory until step 6 promotes them.
    is otherwise silent: the region becomes ordinary prose and the next write appends a second
    one beside it.
 
-6. **Record it, then tighten the checks:**
+6. **Record it. The checks tighten themselves:**
 
    ```bash
    tools/wikitool lint          # unlabelled_edges and unauthorised_labels must be 0
@@ -110,8 +111,14 @@ advisory until step 6 promotes them.
    ```
 
    Only once `lint` reports zero of both is the run finished. The two findings are advisory
-   during the window and become hard errors afterwards - the same path
-   `legacy_citation_markers` took after the citation migration.
+   while `kb_version` is below 4.0.0 and hard from the moment `migrate done` records it -
+   nothing to flip by hand, and no window in which a half-converted corpus is refused by the
+   check that is measuring its progress.
+
+   Do not record the migration to silence the findings. The promotion is what makes the run
+   stick: after it, a bare title in `related:` is a hard error rather than a page still
+   waiting, so a corpus recorded early fails its next lint instead of quietly keeping the old
+   shape.
 
 ## How to tell a migrated page from an unmigrated one