Skip to content

plan(D-AW): base single-sourced + OGAR per-app mint — the shrinkable two-place awareness model #731

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AdaWorldAPI merged 1 commit into from
Jul 18, 2026
Merged

plan(D-AW): base single-sourced + OGAR per-app mint — the shrinkable two-place awareness model #731

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
100 changes: 89 additions & 11 deletions .claude/plans/soa-32-tenant-awareness-redundancy-v1.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,67 @@ whether 64 bits was enough.

---

## §0.5 OPERATOR DIRECTIVE (2026-07-18) — base single-sourced, OGAR mints per-app, two override points

> *"the most important part is making sure that the base layout is only imported
> in one place and ogar responsible for minting per app additional — that way
> activation can be overridden at any time in 2 places and cheaply tested and
> versioned and shrunken later."*

This **supersedes the "add 19 named `ValueTenant` variants" reading of §2.** 19
named base variants would make every M20 shrink a base-layout edit + version bump
+ blast radius across every (currently hand-copied) consumer — the opposite of
shrinkable. The correct shape (which is *also* the already-coded `tenants.md §5`
"facet lane" model — *"a ClassView READING over existing presets, no enum
variant, no layout bump"*):

| Layer | Holds | Override point | Shrink |
|---|---|---|---|
| **Base** — `lance-graph-contract`, ONE import site | reserves a generic **facet region** in the slab (the 324 B headroom, RESERVE-DON'T-RECLAIM) + the `FacetCascade` 16-B shape. Content-blind. | **Place 1** — region size / `ENVELOPE_LAYOUT_VERSION` | region stays reserved, never reclaimed |
| **OGAR** — `ogar-vocab`, per-app mint | mints classid → ClassView whose `value_schema` **projects** the SpoFacet/PearlRung/… *readings* onto the region | **Place 2** — the mint (versioned per app) | **re-mint** the ClassView to drop redundant readings — no base edit, no version bump |

So the A1–A7 + sibling facets of §2 are **OGAR-minted READINGS over one reserved
region, NOT base enum variants.** The base defines *where the bytes are* and
*what shape a facet is*, once; OGAR decides *which readings each app activates*.

**This mechanism already SHIPPED for the triangle (#729 P4, 2026-07-18) — reuse
it, don't reinvent.** `MailboxSoaView::style_rails_at(row, lane) -> Option<[(u8,u8);
6]>` (`soa_view.rs`) reads a 12-byte content-blind register as six `(u8:u8)`
rails — *"one register, two ClassView-selected readings — identical bytes,
different interpretation"* (the operator's "one register, two readings" ruling:
*"226 ARE the frozen; anything else needs 6×2×8bit … for an Orchestration for v3
substrate replayability"*). The **per-class doctrine** (#729 `3248bd9`): *"the
reading is ClassView-selected PER ROW/CLASS, never per lane."* The awareness
facets are **additional `style_rails_at`-shaped 6×(8:8) ClassView readings**,
OGAR-minted per app — the same substrate, one rung wider. `style_lane_at` (12×u8
→ 226-atom FROZEN palette256, `cognitive_palette.rs`) is the FROZEN sibling
reading; awareness facets are the "anything else" orchestration half.
Consequences that gate everything below:

1. **Single-import enforcement (Place 1 integrity).** The base type has one root —
**`lance-graph-contract`** — consumed by exactly the OGAR-aligned cluster:
**`lance-graph-ogar` + `OGAR` + `ogar-vocab`** (the mint/bridge; OGAR/ogar-vocab
carrying the encoder is CORRECT, they are the mint side, not a violation).
Everyone else *imports* the type, never copies. Real violation to close =
ENTROPY **M21**: the non-sanctioned consumer hand-copies (`woa-rs/erp/canon.rs`,
q2 cpic + fma) → one `canon-node-bytes` all import. **`symbiont` is NOT a fix
target** — it is surrealdb-based and **deprecated in favor of OGAR**; its
`bridge.rs:30 NODE_ROW_STRIDE = 512` copy retires *with* symbiont (dropping the
INTEGRATION-PLAN W2c SurrealDB arm from the live path), not by wiring it to the
base. Closing M21 is the prerequisite — a shrink is only "2 places" if the base
is genuinely one root.
2. **The per-app mint moves to OGAR (Place 2 integrity).** The awareness
`value_schema` projection is an **OGAR mint** (classid → ClassView), not a new
hardcoded `BUILTIN_READ_MODES` preset baked in `canonical_node.rs`. The code's
Comment on lines +95 to +96

@chatgpt-codex-connector chatgpt-codex-connector Bot Jul 18, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Keep facet projections off value_schema

This plan makes the per-app awareness projection a ClassView value_schema, but the current contract's ClassView::value_schema returns only the ValueSchema enum preset (Bootstrap/Cognitive/Compressed/Full), and ReadMode.value_schema only selects which fixed ValueTenant set materializes (crates/lance-graph-contract/src/class_view.rs:1133, crates/lance-graph-contract/src/canonical_node.rs:1198). As written, an OGAR mint cannot encode different SpoFacet/PearlRung/sibling readings without either adding a new projection axis/API or hardcoding additional presets, so following this plan would leave the intended per-app facet readings unrepresentable.

Useful? React with 👍 / 👎.

own intent already says so (`:888` "read-mode is layered in one level up").
3. **The auditor's LAYOUT-GATED verdict is re-scoped.** It audited the bake-19
path. The new delta is far smaller: reserve ONE facet region (or a small fixed
count of generic 12-B facet slots) + the ClassView reading contract — NOT 19
named `ValueTenant` variants + 19 `Full`-mask entries. This needs a **re-audit
of the region-reservation delta** before any `canonical_node.rs` edit; the 5
blocking items shrink accordingly (the `Full.field_mask().count() ==
VALUE_TENANTS.len()` lockstep becomes "one region descriptor," not 19).

## §1 Why "spread wide, then measure" (the operator-blessed method)

This is not a new pattern — it is the workspace's own collapse discipline:
Expand All @@ -63,7 +124,14 @@ Either way the conjecture is overcome by measurement, not assertion.

---

## §2 The honest full-width spread — 13 → 32 tenants (derivation, not invention)
## §2 The honest full-width spread — 13 → 32 readings (derivation, not invention)

> **Read through §0.5.** The A1–A7 + sibling table below is the **catalogue of
> OGAR-minted ClassView READINGS** over the base's one reserved facet region —
> NOT 19 named base `ValueTenant` variants. "13 → 32" counts the *readings a
> fully-activated app projects*, not 32 base enum entries. The byte arithmetic
> below sizes the **reserved region**; the per-reading activation is the OGAR
> mint (Place 2).

Current `VALUE_TENANTS` (canonical_node.rs:904-993) = **13 lanes**, offsets
[32,188), 156 B of the 480-byte slab. `ValueSchema::Full` ends at 188;
Expand Down Expand Up @@ -139,16 +207,26 @@ classid + chess `0x06` (dtri1-classid-mint-spec-v1) — never a solo mint.

## §4 Sequencing + gates (nothing solo, nothing silent)

1. **Envelope-auditor gate FIRST.** The A1–A7 + sibling lanes are a layout delta →
`v3-envelope-auditor` reviews the field-isolation matrix, RESERVE-DON'T-RECLAIM,
version-stability *before any canonical_node.rs edit*. (This turn.)
2. **Batched OGAR mint, never solo.** The awareness-facet classids ride the next
batched mint with BoardAggregates + chess `0x06` (dtri1-classid-mint-spec-v1),
not a separate train.
3. **Lane at a time, jc-cert per lane.** A1 (`SpoFacet`) first — it is the user's
own established base design (3 SPO + 3 episodicwitness) and the least
speculative. Each subsequent lane lands with its own facet-layout + a jc-cert
fixture reading real node data.
0. **Single-import prerequisite (Place 1 integrity — §0.5.1).** Base root =
`lance-graph-contract`, consumed by `lance-graph-ogar` + `OGAR` + `ogar-vocab`.
Close ENTROPY **M21** (the woa-rs/q2 consumer hand-copies → one
`canon-node-bytes` all import). `symbiont` is deprecated (surrealdb → OGAR) and
is NOT a fix target — its copy retires with symbiont. A shrink is only "2
places" if the base is genuinely one root. Gates the plan, not just a lane.
1. **Region re-audit FIRST.** The delta is now *reserve ONE generic facet region*
(or a small fixed count of generic 12-B slots) + the `FacetCascade` shape — NOT
19 named `ValueTenant` variants. `v3-envelope-auditor` re-reviews THIS smaller
delta (RESERVE-DON'T-RECLAIM, version-stability, region contiguity) *before any
canonical_node.rs edit*. The prior LAYOUT-GATED verdict audited the superseded
bake-19 path.
2. **OGAR mints the per-app readings, never solo (Place 2).** The awareness
ClassView `value_schema` projections are minted in `ogar-vocab` (classid →
ClassView), riding the batched mint with BoardAggregates + chess `0x06`
(dtri1-classid-mint-spec-v1) — NOT a new hardcoded `BUILTIN_READ_MODES` preset.
3. **Reading at a time, jc-cert per reading.** A1 (`SpoFacet`) first — the user's
established base design (3 SPO + 3 episodicwitness), least speculative. Each
reading lands as an OGAR-minted ClassView projection + a jc-cert fixture over
real node data. Shrink later = re-mint (drop redundant readings), no base edit.
4. **CausalEdge64 / EpisodicEdges64 stay.** No shrink, no delete, until §3's
measured width either vindicates or replaces the 64-bit hypothesis (M20
residual-role ruling, still [H]).
Expand Down