docs(tasks): decompose wave 7 — the ops surface, the alkcall-backed op family behind the default-off feature

This commit is contained in:
glm-5.3-flash committed 2026-10-12 23:28:25 +00:00
1 parent e84a874715
commit c9829836b1
8 files changed
+852 -2

No files matched your search

+2 -2
View File
@@ -1,6 +1,6 @@
---
status: draft
last_updated: "2026-10-12 (wave 6 decomposed into tasks/)"
last_updated: "2026-10-12 (wave 7 decomposed into tasks/)"
---
# alkblobs — Implementation plan
@@ -54,7 +54,7 @@ ops payload surfaces; release readiness is last.
| [4](#wave-4--postgres-kv-engine) | postgres kv engine: schema, pool, CAS rows, stable-cursor list, full engine-state clause families, suite column | wave 2 | decomposed (2026-10-12) |
| [5](#wave-5--pg-lo-large-engine) | pg-lo large engine: LO staged put, descriptorless lo_get windows, companion rows, window-family seam clauses, suite column | wave 4 | decomposed (2026-10-12) |
| [6](#wave-6--fleet-gc) | Fleet GC mechanism: DB-backed pins, joint entry+pin tx, sweep lease, SQL delete arbitration, fleet-validity predicate + constructor declarations | waves 4 + 5 | decomposed (2026-10-12) |
| [7](#wave-7--ops-surface) | `ops` feature: alkcall-backed op family, ACL mapping, pin tokens + renewal, fanout seam | wave 6 | pending |
| [7](#wave-7--ops-surface) | `ops` feature: alkcall-backed op family, ACL mapping, pin tokens + renewal, fanout seam | wave 6 | decomposed (2026-10-12) |
| [8](#wave-8--fuzzing) | Fuzzing: family pattern targets (key encoding, ops payloads), corpus-replay gate | wave 7 | pending |
| [9](#wave-9--release-readiness) | Crate docs, feature-graph verification, README, publish prep | wave 8 | pending |
+85
View File
@@ -0,0 +1,85 @@
---
id: ops-call-harness
name: the in-process alkcall harness — the serving assembly and the client side over one duplex pair
status: pending
depends_on: [ops-foundation]
scope: narrow
risk: low
impact: component
level: implementation
tags: [wave-7, ops, testing, harness, alkcall]
---
## Description
The test harness every ops handler task exercises its handlers
through: an in-process alkcall **serving side + client side over one
`tokio::io::duplex` pair** — the embedder's "dials and hands an
established `Connection` in" shape (ops-surface.md §Why alkcall;
alkcall is a pure protocol crate), realized in-process so the ops
surface is testable server-less.
- **The serving assembly** (the embedder-assembly pattern realized in
test): `OperationRegistryBuilder` receiving the ops family's
spec/handler pairs → the dispatch serving stack (`CallAdapter` /
`GatewayDispatch`, plus `ChannelsAdapter` for the binary ops'
`channel_open` path) → the dispatch run loop over a
`Connection::from_bidi` half (alkcall ADR-009's public stream
constructor). The assembly's shape is the module's documented
embedder entry — the "module exports spec/handler pairs; the
embedder assembles" placement (ops-surface.md's placement note,
the alkgit `git/repo/*` pattern).
- **The client side**: `CallClient` + `spawn_dispatch` over the
duplex pair's other half; `ChannelClient` (`from_connection`,
`open_channel`/`call_open_op`) for the channel ops.
- **The identity/ACL wiring**: a test `IdentityProvider` +
`OwnershipProvider` — the embedder-registry stand-in (namespace →
identities → granted actions) — so ACL behavior is exercisable from
the first handler task onward, including the closed-by-default
denial path (OQ-09).
- **Proven with a stub op**: a trivial stub op registered through the
same path round-trips (call → response envelope), and the channel
path opens with bytes flowing — the pipe proven before any real op
rides it. The stub lives in the test tree; it registers nothing
into the shipped module surface.
## Acceptance Criteria
- [ ] The harness lives in the ops module's test tree
(`#[cfg(test)]` under feature `ops`), parameterized over the
assembled registry (ops register; the harness serves)
- [ ] Server + client halves over one duplex pair; the dispatch run
loop serves; the client calls and receives envelopes
- [ ] The identity/ownership wiring exercisable: a stub op carrying
an `AccessControl` denies an ungranted identity and admits a
granted one (the ACL path proven end-to-end pre-handler)
- [ ] The channel path proven: a stub `channel_open` op establishes
(the `Establishment` phase) and bytes flow both directions
- [ ] `cargo test --features ops` green; the default build unaffected
(the harness is feature-gated test code);
`cargo clippy --all-targets -- -D warnings` and
`cargo fmt --check` green on both shapes
## References
- docs/architecture/ops-surface.md (§Why alkcall — the Connection
hand-in; the placement note)
- docs/architecture/decisions/001-substrate-posture-and-ops-placement.md
§2 (the machinery inventory this harness assembles)
- alkcall (upstream checkout `/workspace/@alkdev/alkcall`):
`src/registry/registration.rs` (`OperationRegistryBuilder`, the
handler kinds), `src/protocol/adapter.rs` (`CallAdapter`),
`src/protocol/connection.rs` (`Connection::from_bidi`),
`src/channels/adapter.rs` (`ChannelsAdapter`),
`src/channels/client.rs` (`ChannelClient`), ADR-047/049 (the
channel_open marker + establishment)
- tasks/ops-foundation.md (the context the harness's assembled
handlers capture)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion
+115
View File
@@ -0,0 +1,115 @@
---
id: ops-e2e-flows
name: the ops wire-level lifecycle proofs — have/need diffing, the token lifecycle, and the ACL matrix end-to-end
status: pending
depends_on: [ops-json-ops, ops-fetch, ops-put]
scope: broad
risk: medium
impact: phase
level: implementation
tags: [wave-7, ops, e2e, tokens, acl, fleet]
---
## Description
The capstone proofs through the harness (`ops-call-harness`) — the
full lifecycle flows the individual handler tests exercise only in
pieces. This is wave 7's equivalent of the engine waves' suite
columns: the surface proven as a whole, on the wire.
- **The have/need flow**: a consumer diffs `blobs/have` over its
digest set → fetches the absent digests (the need announcement *is*
the fetch miss — the diff is caller-side; the wire carries only
concrete fetch requests). Proven end-to-end: seed a store via the
facade, diff from the client side, fetch exactly the absent set,
re-diff to all-present.
- **The fleet token lifecycle end-to-end** (against the harness
server — env-driven, skipping cleanly server-less; the
pglo-poc convention): put → token → renew via have's renewal form
before expiry (`present: true`; the row's expiry observed moving) →
the embedder's registration confirms → the row's protection hands
to the registered liveness source (the namespace binding's
converter) → the row expires naturally and the entry survives on
the registered source's protection. Then the failure half: pause
past TTL → the renewal probe answers `present: false` → the
sweeper reaps the row → the **re-put backstop** (delete-then-
recover; CAS makes the re-put idempotent) → **the race resolves to
a re-put, never to silent loss** (ADR-010 §4's expiry semantics
verbatim, on the wire).
- **The foreign-token refusal**: a token minted by node A presented
to node B (a second fleet instance's ops module over the same
pool) → `present: false`, not an error — the validity domain is
the minting node (ADR-012 §2).
- **The non-fleet token lifecycle**: put → token (the registry holds
the guard) → have-present answers from the custody → the embedder's
registration confirms → the guard converts (the namespace
binding's converter) → the token retires. Proven on a non-fleet
store (sqlite or mem — server-less).
- **The ACL matrix end-to-end**: an unlisted namespace denies read
AND write (closed-by-default — OQ-09); a read-granted identity
fetches/stats/haves but cannot put; a write-granted identity puts
but a read-only peer cannot fetch its namespace; delete only under
the internal authority context (never a namespace grant).
- **The re-routing posture's serving half** (ADR-008 §3): a large
(pg-lo) entry served through `blobs/fetch` from the storage node's
dual-tier fleet store — the shape client nodes in kv-only mode
consume; no fleet-specific op exists (the module supports the shape
without change).
- **The no-side-door audit** (structural, ADR-001 §4): the ops
module's store access enumerated — the public facade plus the
foundation's crate-internal renewal seam only; no `Backend` touch,
no engine types, no path around the facade. Recorded as the audit
the wave-7 review gate re-checks.
- **The renewal-cadence boundary honored** (ADR-013 §1): the tests
drive renewal explicitly (the embedder's cadence role played by the
test); no in-crate renewal worker exists.
## Acceptance Criteria
- [ ] The have/need flow proven end-to-end (diff caller-side; the
wire carries only concrete fetches)
- [ ] The fleet token lifecycle end-to-end vs the harness server:
renew-before-expiry (the row's expiry moves), the
registration-handover, pause-past-TTL → reap → the re-put
backstop — never silent loss (env-driven; skipping cleanly
server-less; acceptance requires green against the server)
- [ ] The foreign-token refusal: `present: false`, not an error
- [ ] The non-fleet lifecycle: custody → confirmation → conversion →
token retirement (server-less)
- [ ] The ACL matrix end-to-end (closed-by-default read+write denial;
the read/write split; delete internal-only)
- [ ] The re-routing posture's serving half: a large entry fetched
through the ops surface from the dual-tier fleet store
- [ ] The no-side-door audit recorded (the store-access enumeration)
- [ ] Gates green: `cargo test` (default), `cargo test --features
ops`, `cargo test --all-features` (the fleet rows live at
acceptance), clippy/fmt across the variants, `cargo doc
--no-deps` clean
## References
- docs/architecture/ops-surface.md (§Op family — the hand-over
contract; the need-half note; §Store-facing relationship — the
re-routing posture)
- docs/architecture/decisions/010-backend-io-seams-and-fleet-state-home.md
§4 (the expiry semantics; the re-put backstop)
- docs/architecture/decisions/012-pre-decomposition-consistency-rulings.md
§2 (the validity domain; the non-fleet generalization)
- docs/architecture/decisions/008-trait-size-probe-fleet-gc-and-large-engines.md
§2/§3 (the TTL/3 cadence; the re-routing topology)
- docs/architecture/decisions/013-family-composition-reactive-boundary-and-feature-graph.md
§1 (the cadence is embedder composition)
- docs/architecture/open-questions.md OQ-09 (closed-by-default)
- tasks/ops-foundation.md, tasks/ops-call-harness.md,
tasks/ops-json-ops.md, tasks/ops-fetch.md, tasks/ops-put.md (the
surface this proves as a whole)
- tasks/suite-column-fleet.md (the env-driven harness-server
convention this task's fleet rows follow)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion
+101
View File
@@ -0,0 +1,101 @@
---
id: ops-fetch
name: blobs/fetch — the verified Sub op and the fanout seam's consumer side
status: pending
depends_on: [ops-foundation, ops-call-harness]
scope: moderate
risk: medium
impact: component
level: implementation
tags: [wave-7, ops, fetch, fanout, verification]
---
## Description
`blobs/fetch` (ops-surface.md §Op family) — the server→client
streaming op: `{namespace, digest, ranges?}` in; **verified bytes
out** over the binary data channel.
- **The op shape**: `OperationType::Sub` (alkcall ADR-021's streaming
handler kind) with the **`channel_open` marker** (alkcall ADR-047)
— the binary data plane rides a `channel_open`-allocated channel;
the establishment phase (alkcall ADR-049) carries the fetch's ack
shape in the `Establishment` reply fields as the wire requires. The
**read** action (ACL exactly like stat/have — closed-by-default).
- **Verified streaming** (ADR-006; ops-surface.md): the *consumer*
hash-checks the received whole against the carried digest —
whole-blob verification; there are no per-chunk or per-range checks
derivable under the canonical digest. The handler streams from the
facade only: `Store::get` (the whole) or `Store::read_range` (the
ranged form). **Ranged fetch** (`ranges?`): the response carries
the same ADR-006 slice-digest convention `read_range` returns —
SHA-256 over exactly the returned bytes, out-of-band.
- **The fanout seam's consumer side** (ops-surface.md; POC #3 finding
A3): the fetch handler consults the foundation's fanout registry —
a live broadcast for the digest → a **subscriber arm** taps it; no
live broadcast, or one that ended (a **late joiner**) → the normal
post-commit verified `get` — identical bytes under CAS. The
slow-subscriber policy (drop-and-late-join) is ops-layer policy,
not a store concern. Named design point: **the producer side's
shape** — the put handler's stream reader as the canonical producer
(serve-while-persisting, A3's shape), and whether fetches also
produce read-side broadcasts (concurrent fetchers of one stored
digest sharing one store read) — decide and record against the A3
text.
- **The miss shape**: a fetch of an absent digest is the typed
missing-error schema (the need announcement *is* the fetch miss —
the have/need diff is caller-side; no separate need op exists).
- **Tests through the harness**: the whole-blob verified round-trip
(byte-exact; the client's hash-check passes); the ranged form
(bytes + slice digest matching `read_range`'s convention); the
fanout consumer side (a live broadcast tapped — the subscriber
receives while the producer runs; a late joiner degrades to the
get path — asserted by which path served); the miss shape; the ACL
denial (no read grant on the namespace → denied).
## Acceptance Criteria
- [ ] fetch exported as a Sub spec/handler pair with the
`channel_open` marker + ALPN; the establishment reply fields
carry the specced ack shape
- [ ] Whole-blob verification: the received whole hash-checks against
the carried digest (the consumer-side discipline documented on
the op; the server streams verified-store bytes)
- [ ] Ranged fetch carries the ADR-006 slice-digest convention
(out-of-band, over exactly the returned bytes)
- [ ] The fanout consumer side: a live broadcast → subscriber arm; a
late joiner → the verified get path; the producer-side design
point decided and recorded
- [ ] The miss shape is the typed missing error; the have/need diff
stays caller-side (no need op invented)
- [ ] ACL: the read action; closed-by-default verified by test
- [ ] Handlers touch only the store facade (no backend side-door)
- [ ] Harness tests green; `cargo test` (default) green;
`cargo test --features ops` green; clippy/fmt green on both
shapes; `cargo doc --no-deps` clean
## References
- docs/architecture/ops-surface.md (§Op family — the fetch bullet +
the fanout text; the need-half note)
- docs/architecture/decisions/006-verification-posture-and-transfer-encoding.md
(whole-blob verification; the slice-digest convention)
- docs/architecture/decisions/008-trait-size-probe-fleet-gc-and-large-engines.md
§3 (the re-routing topology this op serves — large-blob access
through a storage node)
- docs/research/poc-largeblob-findings.md finding A3 (the fanout
seam: one reader, store arm + subscriber arms; late joiners
degrade to get-after-commit)
- alkcall (upstream checkout): ADR-021 (streaming handler for
subscriptions), ADR-046/047/049 (the sink kind, the channel_open
marker, the establishment phase), `src/channels/client.rs`
- tasks/ops-foundation.md (the fanout registry's mechanics),
tasks/ops-call-harness.md (the harness)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion
+160
View File
@@ -0,0 +1,160 @@
---
id: ops-foundation
name: the ops module's substrate — the token layer, the namespace binding, and the fanout registry
status: pending
depends_on: []
scope: moderate
risk: medium
impact: component
level: implementation
tags: [wave-7, ops, tokens, fanout, namespaces]
---
## Description
The `ops` feature's state substrate in `src/ops.rs` (currently the
ADR-001 stub doc comment) — everything the five op handlers ride, and
none of the handlers themselves (they land in the following tasks).
The module is a **store-facade consumer** (ADR-001 §4): its store
access is the public facade plus the crate-internal renewal seam named
below — no backend side-door, no engine types.
- **Module structure + the ops context**: the feature-gated module
shape (`#[cfg(feature = "ops")]`, re-exported from `lib.rs` per the
one-module-per-file convention); the ops-layer context type binding
the `Arc<Store>`, the module's state (the token registry, the
namespace bindings, the fanout registry), and the module config.
This context is what the handler pairs (later tasks) capture.
- **The pin-token layer** (ADR-010 §4; ADR-012 §2): opaque token
minting over a landed put's pin; the **token↔(key, owner) in-memory
mapping with process lifetime** (outlives the minting handler —
ADR-012 §2's correction of the original handler-lifetime wording);
the **validity domain is the minting node** — the registry is
per-process state, there is no cross-node form, and fleet peers
refuse tokens they did not mint (`present: false`, not an error).
Named design points: **the token's minting form** (a CSPRNG id —
e.g. `getrandom` gated behind `ops`; a counter or derivable form is
rejected as a guessable capability — decide and record), and **the
mapping's retention discipline** (when a mapping retires:
conversion-confirmed or process end; the mapping carries no expiry
machinery of its own — the pin row's expiry is the fleet liveness
signal, and nothing expires non-fleet).
- **The guard-custody design** (the fleet/non-fleet split ADR-010 §4
pins): under fleet, the put handler's `Pin` guard **drops at
response completion** — the pin row's expiry is the only protection
from that moment, which is exactly why the TTL/3 renewal duty
exists (a guard held for the token's whole life would make the TTL
machinery dead code on the ops path); under non-fleet, the token
registry **holds the guard** — the token is the in-process guard's
handle (ADR-012 §2's generalization: sqlite, single-instance
postgres, pg-lo, `local`, `mem` — nothing expires outside fleet
mode) until the embedder's registration confirms and the guard
converts. The custody split is the token layer's core branch.
- **The namespace→liveness-source binding** (ops-surface.md §Op
family, hand-over contract step 3): the embedder's converter seam —
per-namespace registration of the converter the embedder plugs its
registry into ("A `Pin` conversion on the server side (token →
registered liveness source) is the mechanism the embedder plugs its
registry into"). The conversion rides the crate-internal
`RegistryCore::convert` machinery (wave 2). Named design point: **
the conversion trigger** (an embedder-called confirm verb vs a
probe-triggered conversion — decide and record; the hand-over
contract's duty text is the embedder's, the mechanism is this
module's).
- **The crate-internal store-core renewal access** (ADR-010 §4: "the
ops module is in-crate and crosses the seam by construction, the
same way the put handler holds the `Pin` guard"): the renewal
probe's access to `pin_renew(key, owner)` and `pin_check(key)`
through the fleet host — the `pin_check` deliberately-unconsumed
note from the wave-6 review (W6-3) names this consumer. The exact
probe composition (verify-the-mapping-then-renew vs
renew-answers-the-verification) is a named design point against
ADR-012 §2's three-step handler duty. What never happens: a *
consumer* reaching the seam, or a public-API widening for this.
- **The fanout registry** (ops-surface.md's fetch fanout seam; POC #3
finding A3): per-digest broadcast state — one reader, store arm +
subscriber arms; bounded subscriber channels (the backpressure
discipline: a slow subscriber stalls the broadcast, which is
correct for a put — subscriber count bounded, slow-subscriber
handling is policy); the **late-joiner signal** (broadcast ended →
the degrade-to-verified-get path); the **drop-and-late-join**
slow-subscriber policy as ops-layer policy. The producer wiring
lands with `ops-put`, the consumer wiring with `ops-fetch`; the
registry's mechanics land here, unit-tested standalone.
- **The renewal-cadence boundary** (ADR-013 §1): the module ships the
renewal *mechanism* (the have form's handler); the TTL/3 *cadence*
— the putter's renewal worker — is embedder composition above the
facade (scheduler/queues), never in-crate machinery. Documented on
the module, not implemented.
- **Dependency posture**: `serde`/`serde_json` (payload handling) and
the token CSPRNG enter as `ops`-gated dependencies — the base crate
stays lean (convention 8; ADR-013 §4's engine table unchanged —
`ops` pulls alkcall plus these, default-off).
- **Unit tests**: token mint/lookup/retirement and the mapping's
process-lifetime semantics; the custody split (the fleet and
non-fleet forms branch as designed); the namespace binding's
converter invocation; the fanout registry's subscribe /
late-joiner-degrade / slow-subscriber-drop mechanics.
## Acceptance Criteria
- [ ] `src/ops.rs` carries the module structure under feature `ops`:
the ops context, the token layer, the namespace binding, the
fanout registry — no op handlers yet (they land in
ops-json-ops/ops-fetch/ops-put)
- [ ] The token layer: opaque, unguessable minting (the CSPRNG design
point decided and recorded); the token↔(key, owner) mapping with
process lifetime; validity domain = the minting node (a
per-process registry; no cross-node form exists)
- [ ] The guard-custody design decided and recorded: fleet — the
guard drops at response completion (the row's expiry is the
liveness signal); non-fleet — the registry holds the guard
(nothing expires) until conversion
- [ ] The namespace→liveness-source binding: per-namespace converter
registration riding `RegistryCore::convert`; the conversion
trigger design decided and recorded
- [ ] The crate-internal renewal access: the ops module reaches
`pin_renew`/`pin_check` through the fleet host (in-crate seam
crossing; no public-API widening; no backend side-door)
- [ ] The fanout registry: per-digest broadcast, bounded subscriber
channels, the late-joiner degrade signal, the
drop-and-late-join policy point recorded
- [ ] New dependencies (`serde`, `serde_json`, the CSPRNG) are
`ops`-gated; the default build's dependency tree is unchanged
- [ ] Unit tests green; `cargo test` (default) green with the module
compiled out; `cargo test --features ops` green;
`cargo clippy --all-targets -- -D warnings` and
`cargo fmt --check` green on both shapes; `cargo doc --no-deps`
clean
## References
- docs/architecture/ops-surface.md (§Op family — the hand-over
contract; §ACL mapping; §Store-facing relationship)
- docs/architecture/decisions/001-substrate-posture-and-ops-placement.md
§3/§4 (placement; the facade-consumer rule)
- docs/architecture/decisions/010-backend-io-seams-and-fleet-state-home.md
§4 (the token contract: minting, custody, validity domain, the
in-crate seam crossing)
- docs/architecture/decisions/012-pre-decomposition-consistency-rulings.md
§2 (the wire shapes: {token, digest}; the have renewal form; the
validity domain; the non-fleet generalization)
- docs/architecture/decisions/013-family-composition-reactive-boundary-and-feature-graph.md
§1 (the renewal cadence is embedder composition), §4 (the feature
graph)
- docs/research/poc-largeblob-findings.md finding A3 (the fanout seam
shape)
- src/ops.rs (the ADR-001 stub this task fills)
- src/gc.rs (`EngineFleetState::pin_renew`/`pin_check` — the seam the
renewal access crosses; `RegistryCore::convert`)
- src/store.rs (`Store::gc()` — the crate-internal access path)
- tasks/review-wave-6.md (W6-3: `pin_check` deliberately unconsumed —
this wave's renewal probe is the named consumer)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion
+119
View File
@@ -0,0 +1,119 @@
---
id: ops-json-ops
name: the JSON control plane — blobs/stat, blobs/have (+ the token-renewal form), blobs/delete
status: pending
depends_on: [ops-foundation, ops-call-harness]
scope: moderate
risk: medium
impact: component
level: implementation
tags: [wave-7, ops, json, acl, tokens]
---
## Description
The three JSON control ops (ops-surface.md §Op family), each exported
as a spec/handler pair the embedder assembles — riding alkcall's
`OperationSpec` + JSON Schema validation (alkcall ADR-016) and its ACL
machinery at op entry (convention 6: this module never re-implements
an authorization check):
- **`blobs/stat`** `{namespace, digest}` → `{len}` — Query; the
**read** action; the handler calls `Store::stat` (the fall-through
general rule, ADR-012 §6.1); a miss is the typed missing-error
schema (not a `{len: null}`).
- **`blobs/have`** `{namespace, digests[], token?}` → `{present[]}` —
Query; **read-gated** (an existence probe over arbitrary digests is
a discovery surface — gated exactly like fetch, not public);
`present[]` corresponds to `digests[]` by index; content never
traverses this op (hashes only). **The renewal form** (ADR-012 §2):
when `token` is present, exactly one digest accompanies it — the
handler verifies the token maps to that digest's pin row (the
foundation's token layer), calls the crate-internal `pin_renew(key,
owner)` via the foundation's store-core access, and answers
`present[]` as usual; an expired or foreign row answers
`present: false` and no renewal occurs — the putter's backstop is
re-put. Without `token`, the pure existence probe (ADR-008's `has`
discipline) over the facade's `has`. **The non-fleet renewal form**:
the token is the held guard's handle — the probe answers present
from the custody (nothing expires; ADR-012 §2's generalization);
the custody/conversion lifecycle is the foundation's.
- **`blobs/delete`** `{namespace, digests[]}` — Mutation;
**`Visibility::Internal`**, evaluated under the internal authority
context per alkcall ADR-017 (internal calls switch authority
context, never skip ACL): requires the embedder's operator/admin
authority context, **not a namespace grant**; the `namespace` field
exists for audit/triage scoping, not gating. The handler calls
`Store::delete` (the arbitration-then-delete path; a `GcRefuse`
refusal surfaces as the typed error schema). External discovery of
an Internal op returns NOT_FOUND (alkcall ADR-017's visibility
discipline).
- **The ACL mapping** (ops-surface.md §ACL mapping; OQ-09's
resolution): **namespace = resource** — `resource_type:
"blob-namespace"`, `resource_id_path` selecting the payload's
`namespace` field; stat/have = the **read** action; **unlisted
namespaces deny by default** (closed-by-default: no ACL entry ⇒ no
read, no write — a grant per namespace is the deferred cost,
deliberately paid over accidental exposure).
- **Schemas** (alkcall ADR-016): input/output JSON Schema per op +
typed error schemas; the namespace field's syntax (non-empty UTF-8
string, embedder-validated — the minting of namespaces is the
embedder's act); the digest wire form is **hex** (the canonical
digest hex — the same form `blobs/put`'s response returns; parsed
via the key layer's hex ingress, malformed → the typed invalid
error, rejects-never-guesses).
- **Tests through the harness** (`ops-call-harness`): stat round-trip
+ the miss shape; have's index correspondence (present/absent
mixed); the renewal form's three answers (live row renews →
present: true; expired row → present: false, no renewal; foreign
token → present: false); delete under the internal authority
context (an external caller gets NOT_FOUND; an internal
operator-context call deletes); the closed-by-default ACL matrix
(an unlisted namespace denies read; a granted identity reads).
## Acceptance Criteria
- [ ] Three spec/handler pairs exported (stat, have, delete); input/
output schemas + typed error schemas per alkcall ADR-016
- [ ] The ACL mapping: namespace-as-resource with the
resource_id_path; the read action on stat/have;
closed-by-default verified by test (unlisted namespace denies)
- [ ] delete: `Visibility::Internal`, the internal-authority-context
requirement (not namespace-gated), external discovery NOT_FOUND
- [ ] have's renewal form: one digest + token; verify-mapping →
`pin_renew`; expired/foreign → `present: false` with no
renewal; the non-fleet form answers from the held custody
- [ ] The handlers touch only the store facade + the foundation's
crate-internal renewal access (no backend side-door — the
ADR-001 §4 rule)
- [ ] Harness tests green; `cargo test` (default) green;
`cargo test --features ops` green; clippy/fmt green on both
shapes; `cargo doc --no-deps` clean
## References
- docs/architecture/ops-surface.md (§Op family — the three JSON ops;
§ACL mapping)
- docs/architecture/decisions/001-substrate-posture-and-ops-placement.md
§2/§4
- docs/architecture/decisions/012-pre-decomposition-consistency-rulings.md
§2 (the have payload shape + the renewal form's three-step handler
duty)
- docs/architecture/decisions/008-trait-size-probe-fleet-gc-and-large-engines.md
(the `has` discipline the non-token form preserves)
- docs/architecture/open-questions.md OQ-09 (closed-by-default, the
risk-asymmetry rationale)
- alkcall (upstream checkout): ADR-016 (operation error schemas),
ADR-017 (the privilege model — internal switches context, never
skips ACL), `src/registry/spec.rs` (`AccessControl`,
`resource_id_path`)
- tasks/ops-foundation.md (the token layer + the renewal access the
have handler rides), tasks/ops-call-harness.md (the harness)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion
+128
View File
@@ -0,0 +1,128 @@
---
id: ops-put
name: blobs/put — the verified Sink op, the pin hand-over, and the {token, digest} response
status: pending
depends_on: [ops-foundation, ops-call-harness]
scope: moderate
risk: high
impact: component
level: implementation
tags: [wave-7, ops, put, tokens, hand-over, verification]
---
## Description
`blobs/put` (ops-surface.md §Op family) — the client→server streaming
op: `{namespace, digest?, len?}` offer + the byte stream in; the
server verifies against the canonical derivation **before commit**;
the response is **`{token, digest}`** (ADR-012 §2).
- **The op shape**: the Sink handler kind (alkcall ADR-046's
`Pub`/Sink — the client→server streaming shape) with the
**`channel_open` marker** (alkcall ADR-047); the establishment
phase carries the offer ack. The **write** action (ACL;
closed-by-default).
- **Verification before commit** (ADR-006; ops-surface.md): the
server hash-checks the received stream against the canonical
derivation (`H("blob <len>\0" + content)`) before commit. A
**known-length offer** (`len` present) is the encouraged one-pass
path; an **unknown-length offer** rides the store's pre-threshold
buffering path (ADR-003's option (d) semantics — the facade's
`put(None, …)` form, the mid-stream stage flush and the
restart-over-staged-bytes derivation pass riding unchanged). When
the offer carries `digest?`, the derived key is checked against it
— a mismatch rejects before commit (the offered digest is a
claim; the derivation is the authority).
- **The pin hand-over** (ops-surface.md §Pin hand-over; ADR-010 §4):
the put lands pinned (pin-before-publish — the facade's put
returns the `Pin` guard); the handler owns the guard; the custody
split is the foundation's design, realized here:
- **fleet**: the guard **drops at response completion** — the pin
row's expiry is the only protection from that moment (the TTL
machinery stays live on the ops path); the token mints over the
committed row (the token↔(key, owner) mapping, owner = the
serving instance's id);
- **non-fleet**: the token registry **takes custody** of the guard
— the token is the in-process guard's handle; nothing expires;
the guard's fate is the custody lifecycle (conversion on the
embedder's confirmation, or process-lifetime hold).
The response returns `{token, digest}` — **both fields always
present** (ADR-012 §2): `digest` is the stored entry's canonical
digest hex (the unknown-length putter cannot know it beforehand and
needs it for every later op); `token` is the minted opaque handle.
Known-length putters receive the same shape.
- **The fanout seam's producer side** (POC #3 finding A3): the put's
stream reader is the canonical fanout producer — **one read** of
the incoming stream (never polled twice), the **store arm**
persists, **subscriber arms** serve concurrent fetchers of that
digest while the put lands (serve-while-persisting); the wiring
into the foundation's fanout registry. Backpressure: bounded
subscriber channels — a slow subscriber stalls the broadcast,
which is correct for a put (the put must land); the
drop-and-late-join escape is the fetch side's policy.
- **The dedup re-put**: CAS dedup means no new entry — the re-put
renews the pin row (fleet: the joint tx's upsert-renewal of
`(key, owner)`) or re-mints the custody guard (non-fleet); the
re-put backstop's mechanics ride the facade's CAS unchanged.
- **Tests through the harness**: the known-length round-trip (the
response's digest matches the git derivation — the byte-exact
standard); the unknown-length round-trip (the buffering path
through the wire); the verification-failure shape (a
byte-corrupted stream → no commit, the typed error, zero residue
on both tiers); the offer-digest mismatch rejection; the hand-over
custody split (fleet: the guard dropped at response — observable
via the row's expiry path; non-fleet: the registry holds the guard
— observable via have-present); the minted token renews via
have's renewal form (the cross-task contract); the dedup re-put;
the ACL denial (no write grant → denied).
## Acceptance Criteria
- [ ] put exported as a Sink spec/handler pair with the
`channel_open` marker; the establishment reply fields carry the
offer ack
- [ ] Server-side verification before commit; an offer-digest
mismatch rejects before commit; a verification failure leaves
zero residue on both tiers
- [ ] The response is `{token, digest}`, both fields always present;
`digest` is the canonical hex (the unknown-length putter's
later-op handle)
- [ ] The hand-over custody split implemented per ADR-010 §4 (fleet:
the guard drops at response completion — the row is the
signal; non-fleet: the registry holds the guard — nothing
expires)
- [ ] The fanout producer side wired: one read of the incoming
stream, the store arm + subscriber arms (serve-while-
persisting)
- [ ] The dedup re-put renews (row upsert fleet-form / custody
re-mint non-fleet)
- [ ] Harness tests green; `cargo test` (default) green;
`cargo test --features ops` green; clippy/fmt green on both
shapes; `cargo doc --no-deps` clean
## References
- docs/architecture/ops-surface.md (§Op family — the put bullet; §Pin
hand-over for remote puts — the cross-wire contract)
- docs/architecture/decisions/012-pre-decomposition-consistency-rulings.md
§2 (the {token, digest} response; both fields always present)
- docs/architecture/decisions/010-backend-io-seams-and-fleet-state-home.md
§4 (the guard drops at response time — the TTL-machinery-alive
rationale; the non-fleet custody)
- docs/architecture/decisions/006-verification-posture-and-transfer-encoding.md
(verification before commit)
- docs/architecture/decisions/003-backend-contract-and-dispatch.md
(the staged-put contract; the unknown-length path the wire rides)
- docs/research/poc-largeblob-findings.md finding A3 (the fanout
producer shape)
- tasks/ops-foundation.md (the token layer, the custody design, the
fanout registry), tasks/ops-call-harness.md (the harness),
tasks/ops-json-ops.md (the renewal form the minted token rides)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion
+142
View File
@@ -0,0 +1,142 @@
---
id: review-wave-7
name: Wave 7 review gate — the ops-vs-alkcall conformance gate
status: pending
depends_on: [ops-e2e-flows]
scope: narrow
risk: low
impact: phase
level: review
tags: [wave-7, review]
---
## Description
The wave-7 review gate (docs/plans/implementation.md §Review gates:
"ops-vs-alkcall conformance (the machinery ridden, not re-invented);
ACL mapping vs OQ-09's closed-by-default; the pin token contract vs
ADR-010 §4/ADR-012 §2; no backend side-door"), before wave 8
decomposes. The gate reads the implementation against the pinned ADR
text line by line (the wave-1..6 gates' method) — the documents, not
memory.
Gate checks:
- **ops-vs-alkcall conformance — the machinery ridden, not
re-invented**: the JSON control plane rides `OperationSpec` + JSON
Schema validation + typed error schemas (alkcall ADR-016); the
binary data plane rides the `channel_open` marker + establishment
(alkcall ADR-047/049) with the Sub/Sink handler kinds (ADR-021/
046); ACL is alkcall's `AccessControl` at op entry (ADR-011) under
the ADR-017 privilege model; producer/consumer vocabulary
throughout; no parallel framing, validation, or authorization
invented anywhere in the module.
- **ACL mapping vs OQ-09's closed-by-default**: namespace-as-resource
(`resource_type: "blob-namespace"`, the `resource_id_path`
selecting the payload's `namespace` field); the read action on
fetch/stat/have, the write action on put; unlisted namespaces deny
by default (no ACL entry ⇒ no read, no write — verified by test,
not by reading the code); delete not namespace-gated at all —
internal authority context only (alkcall ADR-017: internal switches
context, never skips ACL).
- **The pin token contract vs ADR-010 §4/ADR-012 §2**: the token *is*
the pin (no second pin state); the `blobs/put` response
`{token, digest}` with both fields always present; the have
renewal form (one digest + token; verify-mapping → the crate-
internal `pin_renew`; expired/foreign → `present: false`, no
renewal); the token's validity domain = the minting node (foreign
tokens refused `present: false`, not an error); the mapping's
process lifetime (outlives the handler); the guard-custody split
(fleet: the guard drops at response completion — the row's expiry
is the liveness signal; non-fleet: the registry holds the guard —
nothing expires); the TTL/3 renewal cadence is embedder
composition (ADR-013 §1) — no in-crate renewal worker.
- **No backend side-door** (ADR-001 §4): the ops module's store
access enumerated — the public facade plus the crate-internal
renewal seam only; no `Backend` touch, no engine types, no
ops-internal path around the store facade; the e2e audit's
enumeration re-verified against the code.
- **The fanout seam vs POC #3 finding A3**: one reader, store arm +
subscriber arms; late joiners degrade to a verified post-commit
`get` (identical bytes under CAS); the slow-subscriber
drop-and-late-join policy is ops-layer; the producer-side design
point consistent with the A3 text.
- **The base crate stays lean**: the `ops` feature default-off; the
alkcall + serde + CSPRNG deps ride the gate; the default build's
dependency tree unchanged (convention 8; ADR-013 §4's table).
- **Code-smell / hygiene sweep** (the gate's standing hygiene block):
- **`cargo doc --no-deps` zero warnings** (the PGLO-1 class —
public-module doc links resolving to crate-internal items —
de-link rather than export).
- **The `#![allow(dead_code)]` scope notes match their remaining
scope** — the ops module's notes audited; `pin_check`'s
wave-6 "deliberately unconsumed" note updated to its landed
state.
- **Stale forward references updated** — every "wave 7" note in
the module docs reads the landed state.
- **Family conventions**: no inline comments in code (doc comments
on public API fine), thiserror, tokio-only async, no
`unwrap`/`expect` outside tests, naming, one-module-per-file
re-export shape, no blocking I/O on the async path.
- **Task-description correctness**: each wave-7 task's Summary
matches what actually landed (the gate reads the diffs against
the task texts, not the Notes' claims alone).
- **Gates green**: `cargo test` (default), `cargo test --features
ops`, `cargo test --all-features` (the fleet rows live at
acceptance), the pure builds (`cargo check --no-default-features
--features postgres` / `--features "postgres,pg-lo"`),
`cargo clippy --all-targets -- -D warnings` across the feature
variants (default, ops, all-features, the pg-pure builds),
`cargo fmt --check`, `cargo doc --no-deps`.
Findings are fixed inline where cheap, or recorded as wave-8
amendments where they shape the next decomposition (the fuzzing
wave's target inventory reads this wave's surfaces — a payload-shape
finding lands beside it). The gate's completion is what unblocks
wave 8's decomposition.
## Acceptance Criteria
- [ ] Every gate check above read and verified against the cited ADR
text (not from memory of it)
- [ ] Named design points each decided, recorded, and consistent with
the specs
- [ ] Findings fixed inline or recorded as wave-8 amendments with
owners
- [ ] All gates re-run green after any fixes — including
`cargo doc --no-deps` explicitly (the PGLO-1 class)
- [ ] Wave 8 decomposition unblocked (recorded in this task's
Summary)
## References
- docs/plans/implementation.md (§Wave 7, §Review gates — the
ops-vs-alkcall conformance gate)
- docs/architecture/ops-surface.md (the op family, the ACL mapping,
the hand-over contract)
- docs/architecture/decisions/001-substrate-posture-and-ops-placement.md
§2/§3/§4
- docs/architecture/decisions/006-verification-posture-and-transfer-encoding.md
- docs/architecture/decisions/008-trait-size-probe-fleet-gc-and-large-engines.md
§2/§3
- docs/architecture/decisions/010-backend-io-seams-and-fleet-state-home.md
§4
- docs/architecture/decisions/012-pre-decomposition-consistency-rulings.md
§2
- docs/architecture/decisions/013-family-composition-reactive-boundary-and-feature-graph.md
§1/§4
- docs/architecture/open-questions.md OQ-09
- docs/research/poc-largeblob-findings.md finding A3
- tasks/review-wave-6.md (the gate method; the W6-3 note this gate
audits closed)
- tasks/ops-foundation.md, tasks/ops-call-harness.md,
tasks/ops-json-ops.md, tasks/ops-fetch.md, tasks/ops-put.md,
tasks/ops-e2e-flows.md (the wave's tasks)
## Notes
> To be filled by implementation agent
## Summary
> To be filled on completion