From f679ef8034bcbb56c83ab7afbbc8badd2bf711b6 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Mon, 20 Jul 2026 19:43:47 +0000
Subject: [PATCH] docs: record the DocIr-composition seam (a2ui = DocRenderer's
 fourth adapter)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Aligns a2ui-rs awareness with the OGAR DocIr composition grounding
(#217/#218). The composition ruling generalizes this arc, it does not
compete with it: render_stream.rs already says "one template projection;
serve it as a living screen or re-issue it as a document."

Records the OGAR-side widenings this repo will absorb as rename-follows,
not re-derivations:
- a2ui is DocRenderer's fourth adapter (screen == document, A3) — no
  parallel renderer trait
- the FieldView we consume via render_field_view is the Text-leaf reading
  of the E-ONE-MASK-THREE-PORTS fold; when OGAR widens it to the
  renderer-neutral enum (struct -> FieldRow), the wasm/server/paint tiers
  consume the enum's fold (the paint tier's hand-rolled Skin{Form,Flow}
  widget dispatch becomes the variant match) — T1 holds (skins, not a
  second doc-IR vocabulary)
- ObjectSlot = the resolve_nested L1/L2 addressing made explicit
  (desktop->window->region->widget = the A3 brick + ObjectRef +
  ResolutionMode); behavior still by address (T2)

No code change — the composition types are OGAR-side named gaps. Seam
record only.

Generated with Claude Code
---
 CLAUDE.md | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/CLAUDE.md b/CLAUDE.md
index 8cc51b6..dc03795 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -56,6 +56,42 @@ open W1 decision. **Fail-closed:** a missing/narrow role mask never falls back
 to "emit everything" (`full_for` is a *render* convenience, never an RBAC
 fallback). RBAC happens BEFORE framing — the frame is dumb transport.
 
+## DocIr composition grounding (OGAR #217 / #218)
+
+The DocIr composition ruling (`AdaWorldAPI/OGAR docs/DOCIR-COMPOSITION-LAYER.md`
+#217, grounded in `docs/DOCIR-COMPOSITION-GROUNDING.md` #218) is **this arc,
+generalized — not a competing surface.** a2ui-rs is already the code proof:
+`render_stream.rs` says *"one template projection; serve it as a living screen
+or re-issue it as a document."* Ground rules that follow, so a future session
+absorbs the OGAR-side widenings correctly (they are **expansions, not
+rewrites**):
+
+- **a2ui is `DocRenderer`'s fourth adapter.** A3: *"a screen and a document are
+  the SAME positional projection; `DocRenderer` gains its fourth adapter."* The
+  live surface (this repo) sits beside askama / Typst / the paged renderers on
+  one `DocRenderer` trait — do NOT mint a parallel renderer trait. When OGAR
+  renames the doctrine's `ProjectionRenderer` to `DocRenderer`, this repo
+  consumes that trait.
+- **The `FieldView` we consume is a fold, being widened.** `render_field_view`'s
+  `FieldView` (the `ogar-render-askama` struct: `position/label/predicate/value`)
+  is the **`Text`-leaf reading** of the `E-ONE-MASK-THREE-PORTS` fieldview fold.
+  OGAR is widening it to a renderer-neutral `enum FieldView {Text, Badge, Table,
+  ObjectSlot, Geometry, …}` (the struct becomes `FieldRow`, the `Text` payload).
+  When that lands upstream, `a2ui-wasm` / `a2ui-server` / `a2ui-paint` consume
+  the **enum's fold** — the widget-per-field-type dispatch the paint tier
+  already does by hand (`Skin{Form,Flow}`) becomes the enum's variant match. **T1
+  holds:** the enum variants are render *skins* (the fold), never a second
+  closed doc-IR vocabulary beside `RegionKind`.
+- **`ObjectSlot` = our nested-ClassView addressing, made explicit.** The
+  `resolve_nested` L1/L2 drill-down (`a2ui-wasm`) is the ObjectSlot portal:
+  `desktop → window → region → widget` is `ObjectSlot{target, class_view,
+  wide_field_mask}` recursion (the A3 Klickwege brick + an `ObjectRef` +
+  `ResolutionMode`). Behavior still travels by address (T2); the portal carries
+  a projection, never an inline handler.
+- **No code change here yet** — the composition types are OGAR-side named gaps.
+  This section is the seam record so the widening is a rename-follow, not a
+  re-derivation.
+
 ## Layout + status
 
 | crate | role | status |