research: Wikibase/Wikidata deep dive (entity-statement graph, SPARQL); UC-73-75

SHARD-WP-0003 T2. Structure & native-query far-end: typed knowledge graph
(items/properties, statements = claim+qualifiers+references+rank), RDF
projection + SPARQL (WDQS/Blazegraph) incl. federated SERVICE, opaque stable
Q/P identity (labels-as-annotation), statement-level provenance. UC-73
(typed-graph shard, lossy render), UC-74 (SPARQL + federated query), UC-75
(per-assertion provenance). Enriched UC-34/58/52/24. Marks T2 done.
Feeds SHARD-WP-0002 T12/T16.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

spec/UseCaseCatalog.md

@@ -12,7 +12,8 @@ Promoted from `research/260608-c2-wiki-origins/`,
 `research/260614-logseq-deep-dive/`,
 `research/260614-localfirst-workspaces-deep-dive/`,
 `research/260614-trilium-deep-dive/`, `research/260614-wikijs-deep-dive/`, and
-`research/260614-federated-wiki-deep-dive/`.
+`research/260614-federated-wiki-deep-dive/`, and
+`research/260614-wikibase-deep-dive/`.
 See InfoTechPrimers on coulomb.social for use-case catalog conventions.
 
 ## Conventions
@@ -605,7 +606,10 @@ database **query API** (filters/sorts) — a strong delegation target
 (`research/260614-notion-deep-dive/findings.md` §2, §4). **Logseq** shows the third path:
 Datalog over a **derived** index built from plain files — neither native-DB nor plugin;
 when no engine exists, shard-wiki can build the index itself (UC-63,
-`research/260614-logseq-deep-dive/findings.md` §3).  
+`research/260614-logseq-deep-dive/findings.md` §3). **Wikibase** is the far end:
+**SPARQL** over an RDF projection with **federated `SERVICE`** cross-endpoint joins
+(`research/260614-wikibase-deep-dive/findings.md` §2) — a graph-query tier above
+datalog/filters, and `SERVICE` is itself a query-time federation primitive (UC-74).  
 **Priority:** Later
 
 ### UC-53 — Attach a local-first Markdown vault with a live concurrent native editor
@@ -965,6 +969,52 @@ chorus compose with shards that assert a canonical? Feeds SHARD-WP-0002 T1–T5;
 UC-05/UC-27 (union/chorus), UC-26 (fork).  
 **Priority:** Later
 
+### UC-73 — Attach a typed entity-statement (RDF) knowledge-graph shard
+
+**Actor:** Orchestrator / adapter  
+**Goal:** Attach a **Wikibase** instance as a **typed entity-statement shard** — items and
+properties, statements = claim + qualifiers + references + rank — and project entities to a
+rendered page view (lossy to Markdown) **without flattening away the graph**.  
+**Source:** wikiengines, intent  
+**Notes:** The **structure far-end** beyond Notion DB / XWiki XObjects / Trilium relations:
+a true typed knowledge graph with `somevalue`/`novalue` known-unknowns
+(`research/260614-wikibase-deep-dive/findings.md` §1). Content **is not Markdown** — render
+is a lossy projection (UC-55), so attach as a structured shard and either keep the graph as
+the canonical payload (T12) or offer a rendered view, never a silent flatten. Feeds
+SHARD-WP-0002 T12 (typed-graph page payload), T16 (stable opaque identity).  
+**Priority:** Later
+
+### UC-74 — Graph-query the union (SPARQL) and federate queries across endpoints
+
+**Actor:** Reader / analyst  
+**Goal:** Query graph-capable shards with a **graph query language (SPARQL)** and **federate
+queries across endpoints** (`SERVICE`) — a query-time cross-shard join distinct from
+structural (fork/neighborhood) federation.  
+**Source:** wikiengines, intent  
+**Notes:** SPARQL over the RDF projection (Wikidata Query Service / Blazegraph) is the
+**native-query far-end** above Roam/Logseq datalog and Notion filters (UC-52); `SERVICE` is
+a **query-time federation** primitive beside fedwiki's structural federation (UC-72)
+(`research/260614-wikibase-deep-dive/findings.md` §2). Open: union-level common query vs
+pass-through to graph-capable shards. Feeds SHARD-WP-0002 native-query tiering; relates
+UC-63 (derived index — WDQS is one).  
+**Priority:** Later
+
+### UC-75 — Preserve statement-level provenance (references + rank)
+
+**Actor:** Core orchestrator  
+**Goal:** Preserve **provenance at the assertion level** — references and **rank** attached
+to each *statement*, not just per page or per shard — so contradictory values can coexist
+with a curation signal.  
+**Source:** wikiengines, intent  
+**Notes:** Wikibase attaches **references** (sources) and a **rank**
+(`preferred`/`normal`/`deprecated`) to each statement
+(`research/260614-wikibase-deep-dive/findings.md` §1, §5) — provenance granularity **finer**
+than shard-wiki's per-page model, and the *structured* analogue of fedwiki's chorus (UC-72)
+and "view multiple versions" (UC-27). The page model + coordination journal should **allow**
+sub-page/per-assertion provenance even if MVP records per page. Enriches UC-24; feeds
+SHARD-WP-0002 T12 + provenance model.  
+**Priority:** Later
+
 ---
 
 ## B. Knowledge work and collaboration
@@ -1157,7 +1207,11 @@ edit count, and whether the page has local overlays or diverges elsewhere.
 provenance principle. Xanadu deep dive: provenance is *structural*, not decorative —
 content remembers where it came from and where it is reused (UC-45), built into the
 storage model rather than added after the fact
-(`research/260614-xanadu-deep-dive/findings.md` §4, §8.1).  
+(`research/260614-xanadu-deep-dive/findings.md` §4, §8.1). **Wikibase** pushes provenance
+*finer than the page*: **references + rank per statement** let sourced, even contradictory,
+values coexist with a curation signal (UC-75,
+`research/260614-wikibase-deep-dive/findings.md` §1, §5) — the page model should allow
+sub-page/per-assertion provenance even if MVP records it per page.  
 **Priority:** Later
 
 ### UC-25 — Collaborative glossary and precise naming
@@ -1230,6 +1284,9 @@ CamelCase and `[[free links]]`. Markdown-first link semantics TBD.
 | UC-70 | | | ✓⊞ | | ✓ |
 | UC-71 | | | ✓⊞ | | ✓ |
 | UC-72 | | | ✓⊞ | | ✓ |
+| UC-73 | | | | ⬡ | ✓ |
+| UC-74 | | | | ⬡ | ✓ |
+| UC-75 | | | | ⬡ | ✓ |
 | UC-08 | ✓ | | |
 | UC-09 | ✓ | | |
 | UC-10 | ✓ | | |
@@ -1670,6 +1727,38 @@ above fedwiki's fork-DAG provenance edges. Architecture logged for `SHARD-WP-000
 (T1–T5, T11, T13, T16): journal-as-coordination-journal, story-item write granularity,
 journal-replay as a third merge model, fork entries as identity≠placement provenance edges.
 
+### wikibase mapping
+
+(⬡ UC-73–UC-75 are placed in the **wikiengines** matrix column; lineage = the **Wikibase
+deep dive**, `research/260614-wikibase-deep-dive/findings.md`.)
+
+| Wikibase mechanism (findings §) | Catalog UC |
+|---------------------------------|------------|
+| Items/properties + statements (claim+qualifiers+refs+rank); not-Markdown graph (§1) | UC-73 (new) |
+| RDF projection + SPARQL (WDQS/Blazegraph) + federated `SERVICE` (§2) | UC-74 (new) |
+| References + rank per statement = assertion-level provenance + coexistence (§1, §5) | UC-75 (new) |
+| Typed records → typed *graph* entities (§1) | UC-34 (enriched) |
+| Inter-record relations → typed graph edges with qualifiers (§1, §2) | UC-58 (enriched) |
+| Native query → SPARQL/RDF + federated SERVICE (§2) | UC-52 (enriched) |
+| Provenance → statement/value granularity (§1, §5) | UC-24 (enriched) |
+| Opaque stable Q/P IDs; labels = annotations; identity ≠ label/placement (§3) | links UC-51 / T16 |
+| WDQS = derived SPARQL index over canonical JSON entities via update stream (§3) | links UC-63 |
+| Rank: contradictory values coexist with a curation signal (§1, §5) | links UC-27 / UC-72 |
+
+Note: Wikibase (the engine behind **Wikidata**) is the **structure and native-query
+far-end** of the studied set — a **typed knowledge graph** where a "page" is a set of
+**statements** (claim + qualifiers + **references** + **rank**), not prose, queried with
+**SPARQL** over an **RDF** projection (incl. **federated `SERVICE`** cross-endpoint joins).
+Three things transfer cleanly: **opaque stable identity with labels-as-annotation** (the
+cleanest real instance of identity ≠ placement, T16); **assertion-level provenance** —
+references + rank let contradictory values coexist with curation (the *structured* chorus,
+UC-75/UC-27); and **a derived graph index over a canonical entity store** (WDQS = UC-63 at
+scale). **Boundary recorded:** content is **not Markdown** — attach as a structured shard
+and render lossily (UC-55/UC-73), never silent-flatten the graph; expose SPARQL as a
+graph-query tier (UC-74), not the union default. Architecture logged for `SHARD-WP-0002`
+(T12/T16): typed-graph page payload, opaque identity, native-query tiering (SPARQL +
+federated SERVICE), sub-page provenance model.
+
 ---
 
 ## Open questions
@@ -1726,5 +1815,10 @@ journal-replay as a third merge model, fork entries as identity≠placement prov
     vocabulary (create/add/edit/move/remove/fork) at item granularity, or an abstract op
     set other shards can also emit — and does a **chorus / no-canonical** union (UC-72)
     compose with shards that assert a canonical? (Federated Wiki dive §9.)
+24. Does shard-wiki model a **typed-graph page** natively (entities + statements, UC-73) or
+    always project Wikibase to a lossy Markdown/table view (UC-55) — and is **SPARQL/graph
+    query** (UC-74) a union-level capability or a pass-through to graph-capable shards? At
+    what granularity is **provenance** recorded — per page (MVP) or per statement (UC-75)?
+    (Wikibase dive §8.)
 23. How does shard-wiki **honor/surface a shard's path-based access rules** (UC-06) in a
     projection without re-implementing its ACL engine? (Wiki.js dive §9.)