docs(education): document bounded EDU-011 capability

This commit is contained in:
2026-07-31 12:46:22 +08:00
parent e6202187ec
commit 26fa2dfaa2
4 changed files with 75 additions and 32 deletions

View File

@@ -56,9 +56,10 @@ Education must call public APIs, framework extension points, or events. It must
## Background processing, assets, and operational platform
- **Target owner:** Infra platform plus owning domain modules
- **Public/framework capability to reuse:** Infra Job, Redis MQ, File, locks, idempotency, logging, tracing, Excel utilities, tenant propagation.
- **Education-owned gap:** Verified target primitives exist, but durable claim/lease/heartbeat/retry and malware-scanner equivalence are not proven. Education import/export business state is absent or not verified.
- **Allowed external-module change:** Infra Job/MQ/File/logging/observability plus owning Education/Pay/Mall/CRM handlers; scanner deployment or adapter ownership must be decided.
- **Public/framework capability to reuse:** Public Infra File APIs and framework tenant context; generic locks, logging, tracing, and scheduling remain platform capabilities when exposed through public contracts.
- **Education-owned delivered scope:** EDU-011 owns tenant import asset metadata and import jobs, including the five states `PREVIEW`, `PENDING`, `PROCESSING`, `COMPLETED`, and `FAILED`; lease/heartbeat/expired-lease recovery; bounded attempts; and duplicate safety. V4130 is the only delivered EDU-011 migration. Scanner absence defaults to fail-closed `UNAVAILABLE`; CSV/XLSX preview is metadata-only when no parser is available; execution requires both a clean scan and executable parsed content.
- **Deferred scope:** The export boundary currently defines request redaction only—answers and private fields are excluded—but generates no export file or export job. Production scanner integration, full parser availability, retention automation, dead-letter/operator tooling, partial-row reporting, and legacy asset migration remain deferred.
- **Allowed external-module change:** Education may call public Infra APIs only. It must not depend on Infra DOs, mappers, `ServiceImpl` classes, or implementation packages, and it does not move Education job state into Infra.
## Secondary learning, media, AI, and engagement

View File

@@ -160,6 +160,20 @@ V4080/V4090 conditionally seed the five Education permissions (`education:capabi
The accepted Manual Question Collection slice does not warrant a separate ADR: it applies the already-recorded native-authority, tenant-ownership, lifecycle, optimistic-concurrency, route-gating, and snapshot-preservation decisions to one narrower aggregate, without adding a hard-to-reverse architectural trade-off.
### EDU-011 bounded import/export-assets capability (resolved 2026-07-31)
**Decision:** Education owns tenant import asset metadata and durable import-job business state. Infra remains the owner of generic private-file/platform facilities and is consumed only through public APIs.
1. The import job has exactly five persisted states: `PREVIEW`, `PENDING`, `PROCESSING`, `COMPLETED`, and `FAILED`.
2. Processing uses atomic claim, fenced lease token, heartbeat, expired-lease recovery, bounded attempts, and duplicate-safe tenant command keys.
3. Scanning fails closed. With no configured scanner adapter the result is `UNAVAILABLE`, never implicitly clean.
4. CSV/XLSX may produce metadata-only preview when no parser is available. Such a preview is not executable; execute requires a clean scan and explicit executable parsed content.
5. Export scope is request redaction policy only: answers and private fields must be excluded. No generated export file, export worker, or export-job persistence is delivered.
6. V4130 is the only EDU-011 migration. Production scanner integration, full parser delivery, retention/deletion automation, dead-letter/operator tooling, partial-row reporting, and legacy asset migration remain deferred.
7. Education may use public Infra APIs but may not depend on Infra DOs, mappers, `ServiceImpl` classes, or private implementation packages.
The bounded slice does not require a separate ADR: it keeps domain job state with Education and applies the repository's established public-module-boundary rule without introducing a new platform abstraction.
## Unresolved decisions
1. Which Auth/Profile/extended Learning semantics are replaced by Member/System/Infra versus Education-owned, including vocabulary, leaderboard, stats, trend, feedback, exam dates, notifications, points, and badges.

View File

@@ -1,32 +1,48 @@
# EDU-011 — Content import, export, assets, and scanning
# EDU-011 — Bounded content import assets, jobs, and export policy
- **Status:** blocked
- **Status:** bounded capability delivered
- **Type:** implementation program
- **Phase:** 3 / 6
- **Blockers:** EDU-010, Infra File contract, scanner ownership decision, durable job claim decision
- **Delivered migration:** V4130 only
## Tenant-admin outcome
## Delivered tenant-admin outcome
Administrators can import and export education content through durable, duplicate-safe jobs, with secure files, malware scanning, tenant propagation, audit, retries, and partial-failure reporting.
Education owns tenant-scoped import asset metadata and durable import jobs while reusing only public Infra APIs for private-file operations and other platform primitives. The delivered job lifecycle has exactly five states: `PREVIEW`, `PENDING`, `PROCESSING`, `COMPLETED`, and `FAILED`.
## Reuse
The bounded capability provides:
- Education owns import/export business state and content validation.
- Infra owns File, Job/MQ, locks, logging, and audit primitives.
- Scanner integration sits behind a clear adapter; Education does not implement generic storage or scheduling.
- durable atomic claim with lease token, heartbeat, expired-lease recovery, bounded attempts, and terminal failure;
- tenant-scoped duplicate safety for preview requests and execution, so at-least-once delivery does not create a second logical job or repeat completed effects;
- fail-closed scanning: the default scanner result is `UNAVAILABLE`, and unavailable, infected, or errored scans are never executable;
- CSV/XLSX preview metadata when no parser is available; this reports file/type metadata only and does not claim row parsing or content validation;
- execution only after the scan is clean and the preview explicitly reports executable parsed content;
- an export-request redaction policy that excludes answers and private fields from authorized export requests.
## Acceptance criteria
## Ownership and reuse boundary
- [ ] Preview and execute are distinct states.
- [ ] Jobs use atomic claim/lease/heartbeat/recovery semantics.
- [ ] At-least-once retries are duplicate-safe.
- [ ] File type, size, object key, access, and retention are enforced.
- [ ] Scanning fails closed.
- [ ] Tenant context propagates into asynchronous handlers.
- [ ] Exports redact answers and private fields according to authorization.
- [ ] Partial failures and dead letters are visible and auditable.
- Education owns `education_content_import_asset`, `education_content_import_job`, their business state, duplicate keys, lease/recovery semantics, preview policy, and execution orchestration.
- Infra continues to own generic file storage and platform facilities. Education integrates through public Infra APIs only; it does not depend on Infra DOs, mappers, `ServiceImpl` classes, or private implementation packages.
- V4130 is the only EDU-011 schema migration in this slice. No additional migration, generic scanner platform, generic scheduler, or Infra-internal extension is delivered.
## Deferred scope
- No generated export file, downloadable export artifact, export worker, or export-job persistence is delivered. Only the request-time redaction policy is established.
- No production malware-scanner integration is delivered; the default remains fail-closed `UNAVAILABLE` until an external scanner adapter is configured.
- No full CSV/XLSX parser is promised by the fallback. Without an available parser, preview remains metadata-only and execution is blocked.
- File retention/deletion automation, dead-letter tooling, operator UI, partial-row import reporting, and legacy asset migration/re-scan remain deferred.
## Acceptance record
- [x] Education-owned import assets and durable jobs are represented by V4130.
- [x] Jobs use the five-state lifecycle `PREVIEW`, `PENDING`, `PROCESSING`, `COMPLETED`, `FAILED`.
- [x] Claim, lease, heartbeat, expired-lease recovery, retry bounds, and duplicate safety are defined.
- [x] Scanning fails closed, with default `UNAVAILABLE`.
- [x] CSV/XLSX can return metadata-only preview when the parser is unavailable.
- [x] Execute is blocked unless scanning is clean and preview content is executable.
- [x] Export requests apply answer/private-field redaction policy.
- [ ] Generated export files and the remaining operational features are deferred as listed above.
## Risk and rollback
- **Risk:** High operational and file-security risk.
- **Rollback:** Disable job handlers and preserve job/business state for forward recovery.
- **Risk:** Import processing remains security-sensitive; scanner or parser absence intentionally removes executability rather than degrading silently.
- **Rollback:** Disable import handlers while preserving Education asset/job state for inspection and forward recovery. Do not bypass scan or executable-preview gates.

View File

@@ -104,17 +104,29 @@ _Avoid_: Question publication, snapshot invalidation
Descriptive collection metadata reserved for a future access contract. Its presence does not grant, deny, or prove paid, private, membership, or other entitlement in the current Education model.
_Avoid_: Entitlement policy, authorization rule
**Category Publication State**:
The lifecycle of a tenant-owned Subject-scoped Category: Draft is author-editable and student-invisible, Active is immutable and discoverable, and Archived is terminal and undiscoverable. Category does not classify Questions.
_Avoid_: Question Placement, question category relation
**Content Import Asset**:
Tenant-owned metadata that identifies one private Infra-managed source object for an Education import. Education owns the business reference and checksum metadata, not generic file storage or a public download URL.
_Avoid_: Uploaded file owned by Education, public asset URL
**Practice Blueprint**:
A bounded tenant-owned practice configuration attached to exactly one active current-tenant Content Node or Manual Question Collection. Its route and eligible counts are server-derived; it does not define dynamic filtering, entitlement, or descendant traversal.
_Avoid_: Dynamic query, paid practice policy
**Content Import Job**:
A tenant-owned durable Education job that previews and, only when safe and executable, imports content from one Content Import Asset. Its five states are `PREVIEW`, `PENDING`, `PROCESSING`, `COMPLETED`, and `FAILED`.
_Avoid_: Generic Infra job, export job
**Practice Blueprint Publication State**:
The lifecycle of a Practice Blueprint: Draft is author-editable and student-invisible, Active is immutable and returned only while its target route remains available, and Archived is terminal and unavailable.
_Avoid_: Boolean availability as an independent state
**Import Job Lease**:
A fenced, expiring claim on a `PROCESSING` Content Import Job, identified by worker and lease token and kept alive by heartbeat. Expiry permits bounded recovery; it is not proof that the prior worker stopped.
_Avoid_: Redis lock as job ownership, permanent worker assignment
**Executable Import Preview**:
A preview whose source scan is clean and whose available parser produced executable content. Metadata-only CSV/XLSX preview is informative but not executable.
_Avoid_: Successful upload, clean scan alone
**Unavailable Import Scan**:
The fail-closed default when no scanner adapter can produce a clean result. `UNAVAILABLE`, infected, and errored scan outcomes all block execution.
_Avoid_: Scan skipped, assumed clean
**Export Redaction Policy**:
The request-time rule that authorized Education exports must omit answers and private fields. The current bounded capability defines this policy only; it does not generate an export file or persist an export job.
_Avoid_: Delivered export pipeline, downloadable export artifact
**Protected Answer Key**:
Server-only correctness and explanation data captured for stable scoring of a practice session. It is never included in a Safe Question, Question Snapshot JSON, or pre-submit response.