docs(proposals): add 0003 [draft] — typedwasm.region-imports (L13 cross-module)#112
Merged
Merged
Conversation
…ss-module) Closes #95's third acceptance bullet ("if [draft] proposal 0003 is authored, it tracks back here"). Closes the latent seam flagged in proposal 0001 §"Open questions" #5 and the symmetric gap noted in proposal 0002 §"Open questions" #3. ## What 0003 specifies A third producer-neutral carrier (slotting between 0001's regions/capabilities and 0002's access-sites): typedwasm.region-imports carries per-module table of imported regions (source module name, region name, expected schema). Companion to typedwasm.regions which gains a non-breaking high-bit convention for target_region foreign keys: - 0x00000000..0x7FFFFFFE: local region index - 0x80000000..0xFFFFFFFE: imported region (index = value - 0x80000000) - 0xFFFFFFFF: Scalar sentinel (unchanged) v1 typedwasm.regions consumers see "region_id out of range" on high-bit values and recover via lenient codec; v1.1 consumers interpret the high bit as import-table indicator. ## Wire format Mirrors 0001/0002/0004 wire conventions (LEB128 per field, UTF-8 strings, lenient on truncation): u16le version (= 1) u32le_leb128 import_count for each import: u32le_leb128 producer_module_name_len u8[] producer_module_name (wasm module name) u32le_leb128 region_name_len u8[] region_name (exported region name) u32le_leb128 expected_field_count for each expected field: u32le_leb128 field_name_len u8[] field_name u8 kind (FieldKind: Scalar only in v1) u8 wasm_ty (same enum as regions) u8 nullability u32le_leb128 cardinality ## Spec-side mapping Realises MultiModule.idr's ImportedRegion (line 53), SchemaSub (249), ModuleCompat (318), and the flagship noSpoofing theorem (374). Verifier constructs ModuleCompat from emitted bytes by parsing both modules' carriers and checking SchemaSub expected actual. ## 7 error variants MissingDependentRegions, DuplicateImport, UnresolvedProducerModule, UnresolvedExportedRegion, SchemaImportMismatch, ImportTargetOutOfRange, PointerInImportNotSupportedInV1. ## 6 open questions sequenced (1) Pointer-typed expected fields (deferred to v2 — needs SchemaSub to recurse into pointer targets); (2) cross-module access sites (0002 §OQ3 — same high-bit convention extends naturally; recommend folding in at [review]); (3) cross-module capability grants (0004 §OQ4 — capabilities are name-keyed, may resolve differently); (4) link-time vs verify-time resolution (default: separate verify_link_graph pass); (5) producer module name normalisation (NFC deferred to v2); (6) empty import table semantics (omit section entirely recommended). ## Section structure Standard: Context | Proposal (extension + wire format + spec mapping + versioning + producer/consumer obligations + coordination) | Alternatives considered (B extend regions / C component-model / D wasm-imports / E string-table) | Open questions (6 above) | Acceptance criteria (6 gates mirroring 0001/0002/0004) | Related. ## Why [draft] not [review] Neither producer can emit typedwasm.region-imports today. AffineScript has cross-module verification scaffolding (lib/tw_interface.ml for L10/L13 boundary) but Roadmap C3 (multi-module region emission) is the prerequisite. Ephapax cross-module region emission isn't on the roadmap yet. ## README update Adds 0003 entry to docs/proposals/README.adoc's "Current proposals" table. Removes the "Proposal 0003 ... reserved" note (now occupied). Refs: #34, #50, #95, #106, PR #37 (L13 single-module), PR #76 (proposal 0001 with the deferral), MultiModule.idr (entire spec — ImportedRegion through SchemaMismatch). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
3 tasks
🔍 Hypatia Security ScanFindings: 96 issues detected
View findings[
{
"reason": "Action perpolymath/standards/.github/workflows/governance-reusable.yml@main\n needs attention",
"type": "unpinned_action",
"file": "governance.yml",
"action": "pin_sha",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in boj-build.yml",
"type": "unknown",
"file": "boj-build.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in c5-regenerate.yml",
"type": "unknown",
"file": "c5-regenerate.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in cargo-audit.yml",
"type": "unknown",
"file": "cargo-audit.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in codeql.yml",
"type": "unknown",
"file": "codeql.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in dogfood-gate.yml",
"type": "unknown",
"file": "dogfood-gate.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in dogfood-gate.yml",
"type": "unknown",
"file": "dogfood-gate.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in dogfood-gate.yml",
"type": "unknown",
"file": "dogfood-gate.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in dogfood-gate.yml",
"type": "unknown",
"file": "dogfood-gate.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
},
{
"reason": "Issue in dogfood-gate.yml",
"type": "unknown",
"file": "dogfood-gate.yml",
"action": "flag",
"rule_module": "workflow_audit",
"severity": "medium"
}
]Powered by Hypatia Neurosymbolic CI/CD Intelligence |
This was referenced May 30, 2026
hyperpolymath
added a commit
that referenced
this pull request
May 30, 2026
## Summary Closes [#94](#94) — the only remaining open question on proposal 0001 before `[review] → [accepted]` flip. **Decision: WBool is 4 bytes on the wire** (i32.store/load; 0 = false, nonzero = true). The narrower 1-byte encoding is reserved for a future `WBoolPacked` variant in v2. ## Two changes | File | Change | |---|---| | `docs/proposals/0001-multi-producer-carrier-section.adoc` | New `§"WBool wire width — pinned at 4 bytes"` subsection (~36 LOC) inside `§"Wire format — typedwasm.regions"`. Documents rationale, reserves `WBoolPacked` v2, references the rejected lenient alternative. | | `src/abi/TypedWasm/ABI/Region.idr` | `sizeOf WBool = 4` (was `1`) with inline comment pointing at the proposal text and producer-side evidence files. | ## Why 4-byte, not 1-byte, not mixed-width 1. **Producer reality**: both shipping producers already emit 4-byte `i32.store` for boolean fields (AffineScript `lib/codegen.ml:180-190+:373`; Ephapax `src/ephapax-wasm/src/lib.rs:2937-2944`). Pinning at 4 = zero codegen changes. 2. **Spec self-consistency**: `Region.idr` was the outlier at 1 byte; this PR brings spec into alignment with producer reality. `alignOf = sizeOf` (Region.idr:113) carries through to field-alignment math. No proofs or tests hardcode `WBool=1` — verified by grep. 3. **Verifier simplicity**: single canonical width means no new `MixedWidthBoolean` error variant; `region_byte_size` cross-check is deterministic per schema. 4. **Future flexibility**: a 1-byte `WBoolPacked` (or named-equivalent) remains the v2 escape hatch as a new `WasmType` constructor, NOT a width flag on the existing `WBool` codepoint. Additive, not breaking. ## Rejected: lenient mixed-width The `§A` sketch in #94's recommendation (allow either width per field, validate consistency) is rejected because: - `region_byte_size` becomes producer-choice-dependent — same schema → different `schemaSize` values - Forces a new `MixedWidthBoolean` verifier error variant for marginal flexibility - Complicates cross-module schema agreement: a 1-byte WBool field in one module is structurally incompatible with a 4-byte WBool field of the same name in another module ## Risk assessment - Idris2 `sizeOf` change cascades to `schemaSize` / `computeOffsets` / `alignOf` — none of these is hardcoded against `WBool=1` in any test or proof. `MultiModule.idr`'s `Example` schema uses WBool but only via `SchemaEq`/`SchemaSub` structural relations. - `TypedAccess.idr:51` (`HostType WBool = Bool`) is host-side type mapping; unaffected by wire width. - `crates/typed-wasm-verify` `region_byte_size` validation is currently `MAY cross-check` (not enforced) — when implemented later, the canonical 4-byte rule makes the check deterministic. ## Acceptance update for #94 - [x] Proposal 0001 amended with explicit `WBool` wire width. - [n/a] No `MixedWidthBoolean` error variant (strict resolution chosen). - [x] Forward-pointer to `WBoolPacked` future-extension added. ## What this means for 0001 [review] → [accepted] This was the last load-bearing open question on proposal 0001. After merge: | Acceptance gate | State | |---|---| | Criterion 1 (producer signoff) | ✅ | | Criterion 2 (wire↔spec mapping) | ✅ Appendix A (PR #110) | | Criterion 3 (draft codec) | ✅ PR #107 | | Criterion 4 (draft spec doc) | ✅ PR #108 | | Criterion 5 (cross-repo issues filed) | ✅ affinescript#444 + ephapax#221 | | OQ #94 (WBool width) | ✅ **this PR** | | OQ #95 (region-imports) | ✅ proposal 0003 [draft] (PR #112) | | OQ #96 (capability-grants) | ✅ proposal 0004 [draft] (PR #111) | Proposal 0001 is **fully ready for `[accepted]`** after this lands — owner decision is the only remaining step. ## Test plan - [x] `git grep` confirms no hardcoded `WBool=1` size assertions in proofs or tests. - [x] `sizeOf WBool` only used via `computeOffsets`/`alignOf`/`fieldSize` (derived, not hardcoded). - [ ] Idris2 build passes (CI will verify). - [ ] Cargo build + test passes (no Rust code changes — should be a no-op). ## Related - #94 (this PR closes it) - #106 (acceptance roadmap tracker) - #34 (proposal 0001 umbrella) - ephapax#165 §2-Q1 / affinescript#402 §3-Q1 — paired producer reviews - PR #76 (proposal 0001 original landing) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
hyperpolymath
added a commit
that referenced
this pull request
May 30, 2026
…loses #97) (#114) ## Summary Closes [#97](#97) — addresses both halves of the paired-review meander finding. ## (A) Producer-readiness checklist — Appendix B in proposal 0001 (~120 LOC) Tabulates each carrier section's IR prerequisites, current producer-side status, and roadmap-issue identifier in each producer's tracker where the gap is named: | Carrier | IR prerequisites | Status | |---------|-----------------|--------| | `typedwasm.ownership` | per-fn ownership-kind tracking | **SHIPPING** | | `typedwasm.regions` | region decls + field offsets + stable indices | **NOT YET EMITTABLE** (AS Roadmap C1 deferred; Ephapax dead code) | | `typedwasm.capabilities` | per-fn cap decl + module budget + effect IR | **NOT YET EMITTABLE** (AS Roadmap C2 not started; Ephapax Perform/Handle stubbed) | | `typedwasm.region-imports` (0003) | + cross-module region tracking + wasm-module-name resolution | **NOT YET EMITTABLE** | | `typedwasm.access-sites` (0002) | + per-instr tracking + post-rewrite hook | **NOT YET EMITTABLE** (depends on regions) | | `typedwasm.capability-grants` (0004) | + per-callsite grant tracking | **NOT YET EMITTABLE** (depends on capabilities) | NOTE block surfaces the all-or-nothing implication: **do not emit a section you cannot populate.** Proposal 0002 gains a §"Producer-readiness (cross-reference)" subsection (~28 LOC) pointing to 0001's Appendix B plus three access-sites-specific items: per-instruction tracking, post-rewrite hook discipline, region-index stability across sections. ## (B) Canonical emit ordering — extends 0001 §"Producer obligations" New 5th item specifying the recommended emit order matching the runtime dependency graph: ``` 1. typedwasm.ownership 2. typedwasm.regions 3. typedwasm.region-imports (proposal 0003) 4. typedwasm.capabilities 5. typedwasm.access-sites (proposal 0002) 6. typedwasm.capability-grants (proposal 0004) ``` **Consumers MUST NOT depend on this order** — the verifier reads sections by custom-section name, so any order is parseable. The canonical order exists for cross-producer module comparability (byte-equal carrier sequences for two producers of the same logical module — useful for `cmp` / `diff` / content-addressing). ## Acceptance update for #97 - [x] Proposal 0001 gains §"Producer-readiness checklist" appendix listing the IR prerequisites for each section. - [x] Proposal 0002 gains a one-line cross-reference (expanded to a full subsection with access-sites-specific items) to 0001's checklist. - [x] Proposal 0001 §"Producer obligations" specifies the canonical emit order with a "consumer MAY accept any order" note. - [ ] When AffineScript / Ephapax start their respective Roadmap C1/C2/C3 implementations, those issues cross-reference this checklist. *(Producer-side action, not a typed-wasm gate.)* ## What this closes Together with #94, #95, #96 (resolved in PRs #110/#111/#112/#113), this completes all the documentation-side gates around proposals 0001 and 0002. Both proposals are now fully ready for `[review] → [accepted]` owner decision. ## Test plan - [x] AsciiDoc renders cleanly (cols-tables, lower-alpha sub-list, code-block fences). - [x] Cross-reference link to Appendix B from proposal 0002 uses AsciiDoc anchor format. - [x] No code changes — no Cargo/Idris2 build required. ## Related - #97 (this PR closes it) - #34 (proposal 0001 umbrella) - #106 (acceptance roadmap tracker) - ephapax#165 §3+§6 / affinescript#402 §4+§7 — paired reviews surfacing the gap - affinescript#462 (access-sites codegen issue — referenced from Appendix B) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This was referenced May 30, 2026
hyperpolymath
added a commit
that referenced
this pull request
May 30, 2026
…115) ## Summary Owner-decision close of the proposal-acceptance arc. Both proposals satisfied every acceptance gate this session ([see #106 wrap-up comment](#106 (comment))); this PR applies the status flip. ## Status changes - `docs/proposals/0001-multi-producer-carrier-section.adoc` — `review → accepted` - `docs/proposals/0002-access-site-carrier.adoc` — `review → accepted` - `docs/proposals/README.adoc` — "Current proposals" table updated; NOTE block clarifies files remain in `docs/proposals/` (ADR promotion is a separate file-restructure follow-up). ## LEVEL-STATUS.md updates | Level | Before | After | |-------|--------|-------| | L2 (region binding) | proposal-stage | **YES** (carrier-backed) — `verify_access_sites_from_module` (PR #109); gated `unstable-l2` | | L3–L6 (type-compat, null, bounds, result-type) | proposal-stage | **YES** (carrier-backed, schema half) — regions codec PR #107; per-access enforcement gated on producer codegen | | L15-A/B (capabilities) | proposal-stage | **YES** (carrier-backed) — `verify_capabilities_from_module` (PR #109); gated `unstable-l15` | | L13 cross-module (positive form) | — | proposal-stage — proposal 0003 `[draft]` (new row) | "Open gating items" section rewritten to remove obsolete pre-acceptance gates (all shipped via #107/#109) and enumerate the remaining producer-side prerequisites + [draft] proposal 0003/0004 work. ## What this DOES NOT change - **File locations**: proposals stay in `docs/proposals/`. ADR promotion to `docs/decisions/` (with renumbering — existing ADR 0001 is "adopt-rsr-standard", so proposals 0001/0002 would become ADRs 0002/0003) is a deliberate separate PR to keep the status flip cleanly auditable. - **Code**: no source or test changes. Verifier passes already shipped in #107/#109 behind `unstable-l2` / `unstable-l15` feature flags. - **Producer-side codegen**: still pending per Appendix B prerequisites in proposal 0001. ## Test plan - [x] Status field flipped on both proposals. - [x] README table reflects new statuses with date. - [x] LEVEL-STATUS L2/L3-L6/L15 rows show carrier-backed YES. - [x] Open gating items rewritten (4 items: producer codegen, L13 cross-module, L15-C, ADR promotion). - [ ] CI passes (doc-only changes; no Cargo/Idris2 build risk). ## Related - #34 (proposal 0001 umbrella) - #78 (proposal 0002 RFC) - #106 (acceptance roadmap tracker — full wrap-up comment) - This session's 5-PR sprint that satisfied every gate: #107 (codec), #108 (spec doc), #109 (verifier), #110 (Criterion 2 appendix), #111 (proposal 0004 draft for #96), #112 (proposal 0003 draft for #95), #113 (WBool pin for #94), #114 (producer-readiness for #97). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Closes #95's third acceptance bullet ("if [draft] proposal 0003 is authored, it tracks back here"). Closes the latent seam flagged in proposal 0001 §"Open questions" #5 (cross-module region foreign keys) and the symmetric gap noted in proposal 0002 §"Open questions" #3.
What 0003 specifies
A third producer-neutral carrier (slotting between 0001's regions/capabilities and 0002's access-sites):
typedwasm.region-importscarries per-module table of imported regions: source module name, region name, expected schema (for cross-module schema-agreement verification).typedwasm.regions:target_regionforeign keys gain a high-bit convention for import-table indices:0x00000000..0x7FFFFFFE— local region index0x80000000..0xFFFFFFFE— imported region (index = value -0x80000000)0xFFFFFFFF—Scalarsentinel (unchanged)v1 typedwasm.regions consumers see "region_id out of range" on high-bit values and recover via lenient codec; v1.1 consumers interpret the high bit as import-table indicator. No wire breaking change.
Wire format
Mirrors 0001/0002/0004 conventions (LEB128 per field, UTF-8 strings, lenient on truncation):
Section structure
MultiModule.idrspec already covers it, latent-seam framingtarget_regionextension; wire format; spec-type mapping; versioning; producer/consumer obligations; coordinationverify_region_imports_from_modulepass + 7 error variants (MissingDependentRegions,DuplicateImport,UnresolvedProducerModule,UnresolvedExportedRegion,SchemaImportMismatch,ImportTargetOutOfRange,PointerInImportNotSupportedInV1)verify_link_graphpass); (5) producer module name normalisation (NFC deferred); (6) empty import table semanticsSpec-side mapping
Realises
MultiModule.idr's:ImportedRegion.source(line 54)producer_module_nameImportedRegion.regionName(line 56)region_nameImportedRegion.expectedSchema(line 57)SchemaSub(line 249)ModuleCompat from to imp exp(line 318)noSpoofing(line 374, flagship)CompatCertificate(line 142)SchemaMismatch(line 185)SchemaImportMismatcherror variantWhy
[draft]not[review]Neither producer can emit
typedwasm.region-importstoday:lib/tw_interface.mlhas cross-module verification scaffolding for L10/L13 boundary (readstypedwasm.ownershipcross-module), but Roadmap C3 (multi-module region emission) is the prerequisite.src/ephapax-wasm/is single-module.README update
Adds 0003 entry to
docs/proposals/README.adoc's "Current proposals" table; removes the "Proposal 0003 ... reserved" note (now occupied).Test plan
MultiModule.idrline numbers; sibling proposal links; tracker issue links).mainafter PR docs(proposals): add 0004 [draft] — typedwasm.capability-grants (L15-C) #111 merge (no conflicts).target_regionconvention doesn't break any v1 consumer (existing tests).Related
🤖 Generated with Claude Code