Pular para o conteúdo principal

Core Model Readiness and Stabilization Order

Core validation, identity/reference, export/privacy, metadata, localized conversion, extension, local authoring, authority-boundary and semantic-relationship and remaining-core evidence updated 2026-11-11; all-model source/Studio baselines remain dated 2026-10-31. This is an implementation/evidence work programme, not a lifecycle-promotion decision. All-model inventory, original machine-readable snapshot, validation supplement and current combined evidence index.

Tenth bundle: the eleven remaining core requirements — 2026-11-11​

Twenty-two executed Studio cases add bounded observations for REQ-PRI-001, REQ-PRI-002, REQ-PRI-010, REQ-PRI-011, REQ-PRI-012, REQ-PRI-014, REQ-OBJ-010, REQ-OBJ-011, REQ-DOC-001, REQ-DOC-009 and REQ-DOC-016. Combined coverage is now 66 partial / 0 untested / 0 fully evidenced, within the same 66-requirement subset of four core Drafts. This is not complete normative coverage, a 100% completion percentage, or a readiness claim for all 24 specifications. Validator-only scope remains 25 partial / 41 untested.

The archived 22-case report pins Studio 3a85ebf, proposed in PR #900. The new combined index includes nine separately pinned Studio runs and the existing validator observations. Historical bundle counts below retain their original date and scope.

The suite executes the real model/store, a separate local Node inspection, conservative claim gates, portable root/profile rejection, explicit shallow-history migration, actual HTML/JATS/XTG output, pagination, the production collaboration-document initializer and in-memory Yjs update/awareness handling. Machine-translation proposals leave the source unchanged. Explicit acceptance records provider provenance and the human actor while retaining draft/unassessed translation status. Synthetic translation responses and a synchronous checkpoint port are disclosed.

Two split/merge regressions exposed missing explicit identity maps. The paragraph actions now retain a versioned Studio-owned source/result mapping in detached change-event extensions. Real checkpoints and portable history preserve the map, prior snapshots and the surviving ID. Other continuous-editor and replacement paths remain unassessed.

The full Studio npm test invocation passes 1,035 checks across 24 runner invocations, with no failures or skips. The 22 new cases are included in that total. Production build, focused lint and the model import-boundary gate pass. The evidence runner rejects dirty attested source and records exact source hashes and runtime. Website verification re-executes the validator and checks the archived Studio report; it does not rerun Studio locally.

Independent implementations, authenticated live collaboration/UI acceptance, all AI/editorial authority policies, all adapters and screen-reader testing, all historical migration paths, and integrated deployed-release evidence remain open. The shared standalone validator is not independent implementation evidence. Every mapping remains partial, and all specifications remain Draft.

Ninth bundle: semantic relationships and fallback disclosure — 2026-11-05​

Twelve Studio scenarios add partial observations for REQ-PRI-005, REQ-OBJ-007 and REQ-DOC-015, and extend already-partial REQ-PRI-008. Combined coverage is 55 partial / 11 untested / 0 fully evidenced requirements out of 66. Validator-only scope remains 25 partial / 41 untested.

The archived relationship report pins Studio 3180972, proposed in PR #863. The current index combines eight separately pinned Studio runs with local validator cases. They remain observations of one reference implementation, without an integrated deployed-release or independent-conformance claim.

Portable exchange retains explicit note, citation, cross-reference and source-attribution predicates and target roles. Visible superscripts/proximity do not invent relationships. JATS xrefs resolve to distinct note/reference targets and preserve a structured table; HTML carries note/bibliographic links and a table. DOCX retains native footnote/endnote references and bodies. IDML retains the tested anchored footnote and discloses its ordinary-text endnote fallback.

Seven cases exposed missing loss disclosures. DOCX/IDML now report unsupported structured source, citation and cross-reference relationships. XTG/MIF/SLA/LaTeX disclose ordinary note-list and visual-object projections; EPUB discloses omitted public note bodies and flattened objects. Warnings reach the existing export receipt, which stays incomplete and non-lossless, without copying private note bodies or changing the portable source. No claim covers every role, malformed relationship or adapter.

All twelve cases, the full 978-test Studio run, TypeScript, lint and production build pass. External-editor roundtrip/rendering, complete EPUB notes, every nested object role and browser/native acceptance remain open. Every mapping stays partial. Next bundle: rendering and transport versus committed scholarly revisions. No lifecycle promotion is requested.

Eighth bundle: identity, rights and authorization boundaries — 2026-11-05​

Twelve Studio scenarios add partial observations for REQ-MET-008, REQ-MET-009, REQ-PRI-006 and REQ-PRI-013. Combined coverage is 52 partial / 14 untested / 0 fully evidenced requirements out of 66. Validator-only scope remains 25 partial / 41 untested.

The archived authority report pins Studio 7438e81, proposed in PR #862. The current index combines seven separately pinned Studio runs with local validator cases. The new suite hashes real client/server modules and explicitly discloses an isolated in-memory Prisma persistence port; it is not live database/route acceptance or an independent implementation.

The informative studio-authority-boundaries-0.1 scope executes the actual collaboration policy and ticket service. Portable author/editor/ORCID/ROR and OWNER assertions do not create an account session or replace scoped membership. Open licence and rights claims do not issue tickets. Persisted VIEWER access remains read-only after scholarly metadata changes; document/account scope, revocation, suspension and expiry remain authoritative. Collaboration roles do not imply editorial publication approval.

Missing rights remain absent in portable exchange and JATS output, without an invented licence or copyright holder. Inspection does not promote missing provenance or self-asserted authorship into verified facts. A federated administrator prerequisite helper rejects scholarly/unverified claims and unrelated verified email; the narrowly defined trusted matching-email evidence remains only a prerequisite, not an administrator grant or SSO signature check.

The boundaries pass without a production-code change. All twelve cases, the full 966-test run, production build, TypeScript and focused lint checks pass. Live database/network-route enforcement, all service paths, complete institutional provisioning, legal rights clearance and independent interoperability remain open. Every mapping stays partial. The relationship/fallback follow-up is recorded in the ninth bundle above. No lifecycle promotion is requested.

Seventh bundle: local authoring and editable source attribution — 2026-11-05​

Twelve Studio scenarios add partial observations for REQ-DOC-010, REQ-DOC-011 and REQ-PRI-007. Combined coverage is 48 partial / 18 untested / 0 fully evidenced requirements out of 66. Validator-only scope remains 25 partial / 41 untested.

The archived local authoring report pins Studio 8fffc2d, proposed in PR #861. The current index combines six separately pinned Studio runs and local validator cases. They remain observations of one reference implementation, not independent interoperability or deployed-release certification.

Under the informative studio-local-model-authoring-0.1 observation profile, cases clear the account session and deny Fetch, WebSocket, XMLHttpRequest and Node HTTP/HTTPS request/get entry points. Local creation, title/abstract/body edits, validation, document switching, checkpoint and portable save/reopen complete with no guarded network attempt. This covers the initialized model/store with local dependencies; offline app installation/bootstrap and every possible transport API are not established.

Visible source notices and scholarly source metadata can be corrected or removed. Confirmed source links, origin annotations and document import self-attestation do not pin the first heading or ordinary body. Deleting attributed content cleans active links while retaining historical source evidence for explicit revision restoration. A real ProseMirror empty-heading transaction preserves following text.

Two cases exposed silent edit/delete failures under embedded section/block trees. The candidate now traverses both accepted representations for text changes and targeted subtree removal, retaining structure, identities and the existing relationship cleanup. All twelve cases, the full 954-test run, production build, TypeScript and focused lint checks pass.

Browser/native UI acceptance, actual device persistence/crash recovery, hosted authorization, rights clearance, all specialized relationship editors and independent conformance remain open. A visible notice edit does not authenticate or rewrite a separate source relationship. Every mapping stays partial. The identity/rights follow-up is recorded in the eighth bundle above. No lifecycle promotion is requested.

Sixth bundle: extension and vocabulary preservation — 2026-11-05​

Twelve Studio scenarios add partial observations for REQ-DOC-014 and REQ-MET-015, with further evidence for already-partial REQ-MET-014, REQ-OBJ-013, REQ-MET-018 and REQ-PRI-008. Combined coverage is 45 partial / 21 untested / 0 fully evidenced requirements out of 66. Validator-only scope remains 25 partial / 41 untested.

The archived extension report pins Studio 70c559d, proposed in PR #860. The current index combines five separately pinned Studio runs with local validator cases. All five runs belong to one reference implementation; no independent or deployed-release conformance is inferred.

Five import/disclosure regressions fail before the fixes. OJS import discarded foreign metadata fields, treated unknown Open Science members as localized text and shared retained objects with the launch source. The candidate normalizes only owned fields, retains foreign namespace/profile/version/vocabulary data opaquely, and detaches imported graphs. Known metadata edits, checkpoint history and portable save/reopen preserve the tested future terms, nested arrays, falsy values and embedded identities.

Extension properties do not replace core manuscript identity, title or sections. A future translation profile survives portable exchange with explicitly incomplete inspection; an unsupported frame profile is refused without rewriting source. The recognized frame profile enforces its declared alignment vocabulary. DOCX/ODT now disclose omission of the extension graph through warnings and typed downloadable receipt observations without copying its values.

The selected core metadata review profile explicitly identifies authoring/exchange requirements: authoring permits empty titles, exchange rejects them, missing language is diagnosed and an invented mode is refused. This supplies bounded required-field evidence for REQ-MET-015, not full cardinality or controlled-vocabulary coverage. The twelve cases, full 942-test Studio run, production build and focused lint checks pass.

Generic namespace/version validation, every profile/vocabulary, foreign semantic interpretation, external profile resolution, complete metadata constraints and independent interoperability remain open. Every mapping stays partial. The local authoring/source editability follow-up is recorded in the seventh bundle above. No lifecycle promotion is requested.

Fifth bundle: localized metadata and conversion loss — 2026-11-04​

Twelve Studio scenarios add partial observations for REQ-MET-011, REQ-MET-017 and REQ-PRI-008, with further preservation evidence for already-partial REQ-MET-018. Combined coverage is 43 partial / 23 untested / 0 fully evidenced requirements out of 66. Validator-only scope remains 25 partial / 41 untested.

The archived localized conversion report pins Studio f191c97, proposed in PR #859. The current index combines four separately pinned Studio runs with locally executed validator cases. These are observations of one reference implementation, not independent interoperability or deployed-release certification.

The cases exposed regional locale collapse during OJS normalization and a missing primary-language map entry during localized edits. Recognized language tags retain region/script distinctions, legacy underscore spellings are canonicalized, and conflicting aliases reject the import before returning a candidate. The source launch stays unchanged. Localized editing retains the scalar primary abstract/keywords while seeding their missing localized entry.

Actual DOCX/ODT package XML confirms that secondary localized abstracts/keywords are absent. The candidate now discloses those omissions and missing scholarly metadata through the export warning/report path. Typed observations distinguish preservation, normalization, omission and unverified mapping, without copying metadata values. Empty variants absent from output are disclosed as omitted. Portable save/reopen retains all tested localized variants; conversion never invents missing rights or publication facts.

Seven import/edit cases fail on the preceding implementation and pass after the fixes. The twelve new cases, full Studio test run and production build pass; focused office/export/privacy checks also pass after the final report adjustment. Complete localized titles, all metadata profiles, every adapter, HTML-to-text correctness, full language-tag support and independent round trips remain open. Every mapping stays partial.

The extension/vocabulary follow-up is recorded in the sixth bundle above. No lifecycle promotion is requested.

Fourth bundle: metadata assertion states and provenance — 2026-11-04​

Twelve Studio scenarios add partial observations for REQ-MET-001, REQ-MET-005, REQ-MET-010, REQ-MET-012 and REQ-OBJ-014, with further preservation evidence for already-partial REQ-MET-018. Combined coverage is 40 partial / 26 untested / 0 fully evidenced requirements out of 66; the validator-only scope remains 25 partial / 41 untested.

The archived metadata report binds its exact Studio source commit and file digests to PR #858. The current index combines three separately pinned Studio runs with local validator checks. It records observations at those commits rather than certifying an integrated or deployed release. All three suites belong to one reference implementation.

The cases exposed loss of existing ORCID assertion evidence and private visibility during contributor edits, fabricated registry provenance for typed ROR values, and replacement ROR values treated as verified from their syntax. The candidate preserves unchanged assertion IDs, status, timestamps, sources and visibility; replacement identifiers remain self-asserted without inherited verification, while explicit ROR lookup selection records registry provenance without claiming identity or membership verification. Metadata audit events now name the structural change, its subject, property and value instead of labeling it as an abstract edit.

Assertions, visibility states and inferred source attribution survive checkpoint and portable save/reopen. Inspection does not verify claims or mutate evidence. Verified/rejected fixtures are synthetic preservation inputs, not independent real-world proof. Complete metadata profiles, external identity/membership verification, all import/conversion paths, server authorization and independent interoperability remain open. Every mapping remains partial.

The localized metadata/conversion follow-up is recorded in the fifth bundle above. No lifecycle promotion is requested.

Third bundle: export credentials and confidential metadata — 2026-11-04​

Twelve Studio scenarios add partial observations for REQ-PRI-009, REQ-MET-013 and REQ-PRI-003, with further preservation evidence for already-partial REQ-MET-018. Combined coverage is now 35 partial / 31 untested / 0 fully evidenced requirements out of 66. The original validator-only scope remains 25 partial / 41 untested; requirements are counted once across roles.

The archived export/privacy report pins Studio da5b530, from PR #857, which was open at capture. The current index combines this with the prior identity/reference report and the locally executed validator; it does not certify deployment or independent interoperability.

The cases exposed private-note disclosure in custom HTML/PDF-source/DOCX exports, credential fields in separately written .omi history, and common underscore/hyphen credential aliases bypassing checks. The candidate uses the existing public annotation projection for custom exports and refuses portable credentials across the complete source graph before container construction. The OMI reference validator now recognizes the same named field forms. Errors omit credential values, while scholarly prose and unrelated extension fields remain unchanged.

Thirteen standard public output paths and the three custom paths are inspected. Full authoring backups retain private notes and history; public publication artifacts filter supported private identity and note assertions. Arbitrary secrets in prose/encoded/binary data, all extension audience policies, every export path and server-rendered PDF bytes remain outside this evidence. Every mapping remains partial.

The metadata follow-up is recorded in the fourth bundle above. No lifecycle promotion is requested.

Second bundle: identity continuity and deletion — 2026-11-04​

Twelve Studio scenarios exercise actual edits, formatting, renaming, subtree moves, reparenting/cycle rejection, section layout, deletion, checkpoints, .omi save/reopen and explicit revision restoration. They add seven partial requirement observations: REQ-PRI-004, REQ-OBJ-004/008/009/016 and REQ-DOC-002/003. Combined with the validator evidence, the 66-requirement inventory now has 32 partial / 34 untested / 0 fully evidenced entries. The validator-only manifest retains its separate 25-partial/41-untested scope.

The archived Studio report pins commit b9c80c1, file digests and case mappings. Studio PR #856 was open at this evidence capture; these observations do not establish a merged or deployed release. OMI CI verifies the archived report and re-executes its own validator cases; Studio CI independently runs the same Studio case code, which remains reference-implementation evidence rather than a second implementation.

The tests exposed a save/reopen failure after deleting a referenced figure and missing descendant deletion events. The candidate fixes remove affected active relationships atomically, leave visible unresolved-reference text in surviving prose, retain the original relation in history, record descendant deletion and reject reserved IDs before a store update, including before checkpoint. Strict portable-reference validation remains in force. Explicit revision restoration recovers the semantic objects and relationships.

Full editing-path coverage, all specialized relationships, split/merge equivalence, native/browser acceptance and independent interoperability remain open. The export/privacy follow-up is recorded in the third bundle above. No lifecycle promotion is requested.

First validation test bundle — 2026-11-04​

Twelve new executable behavioral cases cover REQ-DOC-012, REQ-DOC-013, REQ-OBJ-015, REQ-OBJ-018 and REQ-MET-018 in the core snapshot validator role. They exercise frozen input graphs, deterministic located diagnostics, exact resource boundaries, truncated results, cycles, detached reports and inert asset references. The existing 22 JSON fixtures remain unchanged. Validator-only coverage is 25 partial / 41 untested / 0 fully evidenced requirements out of 66. See the scope and exclusions for each requirement.

The validator passed these checks without a runtime change. Editor locks, raw-byte parser limits, full metadata conversion and independent conformance remain outside this bundle. Next bundles are identity continuity, dependent references after deletion, and credentials/confidential metadata in exports. No lifecycle promotion is requested.

Completed in the original 2026-10-31 review​

  • Reconciled all 24 registry entries against source and pinned Studio evidence. Corrected obsolete 0.2.0 runtime claims: Studio already uses the 0.3.0 manuscript/schema boundary.
  • Rewrote OMI-SPEC-000/120/100/140 using the canonical 23-section structure, with 66 stable requirement IDs and explicit scope, compatibility, privacy, accessibility and review gaps.
  • Published a bounded core snapshot projection schema, pure validator, 22 digest-bound fixtures and complete requirement accounting. Twenty requirements have partial fixture evidence; 46 are untested.
  • Prepared a separate Studio review panel that keeps authoring editable, distinguishes authoring/exchange requirements, binds report freshness to source state and preserves existing report and producer contracts.
  • Added local/CI checks for fixture integrity, expected outcomes, deterministic non-mutating results, requirement inventories, registry/source versions and evidence links.

All specification revisions remain Draft. No external review, independent implementation approval or stable release is represented as completed.

Stabilization sequence​

PhaseModelsRequired next evidence
1000, 120, 100, 140Review the new requirements; complete role-specific schema/behavior coverage and compatibility decisions.
2150, 110, 130Identity/assertion schemas, anchor persistence across edits, annotation lifecycle and deletion/retargeting fixtures.
3160, 180Requirement-mapped revisions/checkpoints/tombstones/digests and canonical validation report/profile contracts.
4220, 210, 221, 225Coordinated bibliography, citation, library and derivation review; resolve the 210/221 normative dependency cycle.
5320, 330Full parser/serializer/migration/asset-preservation coverage against a selected pinned file/container profile.
6190, 170, 200Scoped permissions, translation/review semantics, revocation and adversarial anonymity tests across transports and exports.
7240, 230Reproducible source-preserving publication and explicit release/acceptance authority.
8340Per-format mappings, loss reports, negative fixtures and declared round-trip invariants.
9350, then 300/310Approved conformance/capability profiles, plugin/API negotiation and independent interoperability evidence.

The 350 measurement infrastructure is used from the beginning; final normative stabilization follows mature targets. Security, privacy, data-loss prevention and regression fixes remain continuous priorities regardless of phase. The table is a work order, not an instruction to postpone production defects.

Candidate decisions​

Each model needs a public record identifying current/proposed state, exact versions, applicable role/profile, reviewed evidence, unresolved issues and compatibility effects. Progress through Draft, Review Candidate, Implementation Candidate and Stable is separate from deploying working Studio features. The core rewrites have not completed public review and are not automatically Review Candidates.

Concrete remaining work​

For the current core group, next steps are normative review, requirement-by-requirement applicability, additional behavioral fixtures for identity continuity and deletion, semantic/projection preservation, privacy/audience handling, accessible language-aware output, unknown extension round trips and meaningful independent validation. The same validator copied into Studio cannot supply the independence requirement.

The main Studio producer continues to declare its existing exact semantic specification revisions. The separate review profile targets the newly proposed Drafts without retroactively changing saved documents. Any future producer adoption needs compatibility and migration evidence.

Current source limits​

The snapshot pins the specification baseline and Studio main commit, and records PR #785 as open at review time. It does not claim that a successful PR was merged or deployed. Earlier matrix counts mixed dated baselines and newer changes; they should not be used as completion percentages. Re-run consistency checks and refresh evidence after merges or substantive model changes.