Mythenheim uses SemVer with Fluxheim-style release gates. A feature is not done because it exists in code; it is done when it has tests, docs, config validation, security review notes, and a smoke path.
0.x: incubator releases. Schema and APIs may change.1.0.x: stable forum-core bugfixes only.1.x.0: stable additive modules.2.0.0: breaking API, schema, plugin, or security-boundary changes.
Security fixes should be backported to the latest stable minor when practical.
Scope:
- Rust 1.97.1 toolchain pin.
- EUPL-1.2 license and plugin/theme notice.
- Axum health service.
- Config parser and
--check-config. - Safe Markdown rendering primitive with XSS tests.
- Capability string validation.
- GitHub CI, check script, release metadata validation, doc link validation.
- Binary portability CI for Linux, Windows, and macOS.
- Native release binary packaging script and release artifact documentation.
- Rootless Podman SurrealDB test helper.
- Fluxheim proxy notes and Wolfi reverse-proxy smoke fixture.
Done when:
scripts/checks.shpasses.- Native Linux, Windows, and macOS Rust tests pass in CI.
- Local validation release binary build produces a package and SHA256 values.
cargo deny checkandcargo auditare runnable in CI.scripts/smoke_local.shpasses.- Fluxheim Wolfi can proxy
Host: mythenheim.euto Mythenheim/healthz.
Scope:
- SurrealDB migration definitions and CLI validation.
- Rust SDK admission remains deferred until selected feature flags pass the all-feature security policy.
- Schema bootstrap for
user,role,category,topic,post,session, andaudit_log. - Migration runner with idempotency tests.
- Podman SurrealDB integration test script using random ports.
Done when:
- Unit tests cover schema generation and migration ordering.
- Integration smoke creates a temporary namespace/database in SurrealDB.
- Failed migrations cannot partially mark themselves complete.
Scope:
- Password policy, Argon2id hashing, and verification primitives.
- Opaque session token generation, server-side token hashing, and constant-time token hash verification.
- Registration and login.
- Opaque session tokens stored hashed server-side.
- Secure cookie settings and logout revocation.
- Preview HTTP endpoints for registration, login, current-user lookup, and logout.
- Account lockout and login rate-limit hooks.
Done when:
- Password hash tests verify successful and failed verification.
- Session token tests verify random generation, hashed storage shape, and constant-time verification.
- Session revocation tests prove old cookies stop working.
- Auth endpoints reject malformed JSON and oversized bodies.
- Login lockout tests prove repeated bad credentials return
429andRetry-After. - Unknown-login tests prove the auth store initializes a dummy password hash used to reduce account-enumeration timing differences.
Scope:
- Category tree.
- Topic create/read/list.
- Post create/read/edit.
- Markdown to sanitized HTML persistence.
- Slugs, pagination, soft delete primitives.
Current preview:
- In-memory category, topic, first-post, and reply service.
- Public read API and session-required write API.
- Private category reads require an authenticated session.
- Flat and nested category read APIs.
- Direct post read API.
- Post edit revisions and soft-delete primitives for posts/topics.
- Topic list pagination accepts query parameters.
- Sanitized Markdown persistence shape with raw HTML dropped before rendering.
Done when:
- XSS fixture tests cover Markdown, raw HTML, links, and images.
- Permission tests cover public/private category reads.
- API smoke test creates a category, topic, and reply.
Scope:
- Capability resolver.
- Global roles and category-scoped moderator roles.
- Ownership checks such as
post.edit.own. - Trust levels TL0 through TL3.
- Role assignment escalation prevention.
Current preview:
- Pure Rust permission resolver for capability strings, trust-level grants, global roles, category-scoped roles, ownership checks, and assignment escalation prevention.
- Deterministic trust-level calculation from activity metrics.
- Preview forum routes use the permission resolver for category creation, private reads, topic creation, replies, edits, and soft deletes.
- In-memory permission service stores roles, default role grants, global user role assignments, and category-scoped role assignments with escalation prevention.
Done when:
- Tests prove actors cannot grant capabilities they do not hold.
- Category-scoped moderation does not leak across categories.
- Trust promotion/demotion is deterministic and audited.
Scope:
- Report queue.
- Approval queue.
- Warnings and warning points.
- Mute, ban, and shadowban state.
- Append-only audit events for every staff action.
Current preview:
- In-memory moderation service with report and approval queues.
- Preview HTTP routes for post reports, queue reads, warnings, shadowban changes, queue resolution, and audit reads.
- Warning records with active warning points.
- Automatic mute and ban state at configured warning point thresholds.
- Warning expiration that marks warnings inactive, recomputes points, and audits the state transition.
- Transactional moderation macro execution in the preview service.
- Stored moderation macro definitions with preview routes to create, list, read, and execute reusable action lists.
- Delayed moderation jobs that execute due macro action lists once and preserve state on failed jobs.
- Preview HTTP routes to schedule delayed moderation jobs, run due jobs, and read job status.
- Shadowban state and forum read filtering that hides shadowbanned authors from other users while preserving the author's own view.
- Append-only audit events for report creation, approval queuing, warning issuance, and shadowban changes.
- Audit events capture previous and new user moderation state for stateful staff actions.
- Staff-facing moderation routes require explicit capabilities such as
moderation.queue.read,moderation.queue.write,user.warn,user.shadowban,moderation.macro.read,moderation.macro.write,moderation.macro.execute,moderation.job.read,moderation.job.write, andaudit.read.
Done when:
- Shadowbanned users see their own posts while others do not.
- Audit tests prove previous state is captured.
- Warning threshold tests trigger automatic mutes/bans.
Scope:
- Full-text search indexes.
- Precise read/unread state.
- Watched topics/forums/tags.
- Mentions and notification records.
- WebSocket notification delivery preview.
Done when:
- Read-state tests cover new reply and mark-read flows.
- Search tests cover permissions and deleted content.
- WebSocket smoke test receives a mention notification.
Scope:
- Attachment upload metadata and limits.
- MIME and size validation.
- Link/image restrictions for low trust users.
- Keyword/regex filters.
- Rate limiting by IP and user.
Done when:
- Upload tests reject executable and mislabeled files.
- TL0 link/image restrictions are enforced.
- Rate-limit tests cover login and post creation.
Scope:
- Admin settings API.
- Moderator dashboard API.
- Custom moderation macros.
- Delayed moderation jobs.
- Metrics and health detail endpoints.
- OpenTelemetry trace propagation and service spans.
- Prometheus-compatible metrics endpoint.
- OTLP export config compatible with Jaeger or an OpenTelemetry Collector.
Done when:
- Config validation protects unsafe settings.
- Moderation macro tests prove transactional action behavior.
- Job tests prove delayed moderation executes once.
- Prometheus scrape smoke verifies bounded-label metrics.
- Jaeger/OTLP smoke verifies request traces from Fluxheim to Mythenheim.
Scope:
- MiniJinja rendering.
- Theme inheritance and style properties.
- Template modification system.
- CSP nonce/reporting integration.
- Cached compiled templates.
Done when:
- Template tests reject unsafe function/database access.
- Conflicting template modifications are deterministic.
- CSP tests prove inline scripts without nonce are blocked by policy output.
Scope:
- Versioned WIT interface.
- Hook manager.
- Plugin manifest validation.
- Host capability grants.
- Plugin timeout, memory limit, and failure isolation.
Done when:
- Malicious fixture plugin cannot access filesystem/network/database.
- Panic/trap disables only the plugin, not Mythenheim.
- Hook ordering and mutation tests are deterministic.
Scope:
- Import framework for legacy forum exports.
- User, category, topic, post, attachment import mapping.
- Dry-run validation and import reports.
Done when:
- Fixture imports preserve post authorship and timestamps.
- Dry-run produces no writes.
- Permission mapping gaps are reported clearly.
Stable scope:
- Accounts, opaque sessions, password login.
- Category/topic/post core.
- Sanitized Markdown content.
- RBAC/ABAC permissions.
- Trust levels.
- Moderation queues, warnings, bans, shadowbans.
- Audit log.
- Search, read state, watched content, notifications.
- Attachments with strict validation.
- Rootless Podman deployment and direct compiled binary deployment.
- Direct compiled binary support for Linux, macOS, and Windows.
- Fluxheim reverse-proxy compatibility.
- OpenTelemetry tracing and metrics with Prometheus and Jaeger/OTLP smoke tests.
Done when:
- Full release checklist passes.
- Podman and direct binary smoke tests pass.
- Binary portability CI passes for Linux, Windows, and macOS.
- SurrealDB integration tests pass from a random rootless port.
- Fluxheim Wolfi reverse-proxy smoke passes for
mythenheim.euanddev.mythenheim.eu. - Observability smoke proves Prometheus metrics and Jaeger-visible traces.
- Security review has no unresolved high or critical issues.
- Upgrade path from every
0.1xschema preview is documented or explicitly rejected with migration guidance.
Stabilize theme inheritance, style properties, TMS, CSP, and server rendering.
Stabilize plugin APIs, host grants, compatibility checks, and plugin admin UI.
Stabilize WebAuthn/passkeys and OIDC SSO with account-linking tests.
Stabilize RSS/Atom, webhooks, and ActivityPub-compatible federation after abuse controls are ready.
Add clubs/groups, galleries, events, and richer profile/community mechanics.
Add moderation replay mode, workload balancing, and safety budget dashboards.
Done when:
- Replay tests prove simulated moderation actions do not mutate production state.
- Queue-routing tests cover language, category, severity, and conflict rules.
- Metrics tests keep labels bounded and privacy-safe.
Add hash-linked audit exports, per-topic governance modes, proposal voting, expert-only replies, slow mode, and consensus summaries.
Done when:
- Audit-chain verification survives backup/export/import round trips.
- Topic-mode permission tests cover every mode.
- Vote and summary tests prevent duplicate or unauthorized actions.
Add user-owned content packages for posts, attachments, bookmarks, profile data, and redaction-aware exports.
Done when:
- Export tests prove private/deleted/third-party data boundaries are respected.
- Redaction tests prove secrets and other users' private data are excluded.
- Import tests can restore a user-owned archive into a test namespace.
Add plugin capability simulation and theme security linting before production activation.
Done when:
- Plugin simulation tests prove requested grants match actual host access.
- Theme lint tests catch CSP conflicts, unsafe inline behavior, accessibility regressions, and layout overflow.
- Failed checks block activation without breaking the currently active theme or plugin set.
Add privacy-preserving community health metrics for response time, unanswered topics, moderation delay, newcomer success, and content lifecycle transparency.
Done when:
- Analytics tests prove no usernames, emails, IPs, topic titles, post content, or raw user agents become metric labels.
- Lifecycle tests show pending/hidden/rate-limited/edit states to users without exposing anti-spam internals.