Implementation criterion review and operating limits
System criterion assessment
Requirement |
Observable criterion exercised |
Executable cases |
|---|---|---|
Real HTML build resolves cross-page explicit seeds and independent seedless views; mixed unknown/hidden seeds fail at invocation location. |
|
|
Two authored schemas retain host labels and placement; invalid mapping reports its entry path and invocation location before view generation. |
|
|
Generated output explores and loads rich local details under file and nested static HTTP URLs with external HTTP/HTTPS blocked by proxy. |
|
|
Actual host traces traverse in both directions; hidden references are filtered; host broken-link failure coexists with a usable missing node. |
|
|
Software-seeded expansion reveals higher declared levels above it; the added hardware rank and right-side shared test retain geometry. |
|
|
Hover/keyboard details preserve host metadata, rich body, tables, local images and namespaced IDREFs; external metadata and unavailable bodies are explicit; original links and remote-media fallbacks work. |
|
|
Exploration and viewport changes in one view leave the other view’s revision and transform unchanged. Async completion does not reclaim focus from a control outside the view. |
|
|
The real 30,000-Need/90,000-edge build loads and explores to exactly 200 visible nodes; another complete group is rejected without changing the committed revision. Measurements repeat under file and HTTP URLs. |
|
|
A wheel installs offline into a fresh Python 3.10 host and builds the fixture with Node absent from PATH; import resolves inside that host. |
|
|
Interaction workflows use the installed stable Firefox desktop binary on Windows, including pointer, keyboard, pan/zoom, paging and recovery. |
|
Software criteria and transition review
The individual acceptance criteria are reproduced beside their executable cases in Executable requirement coverage. Source review checks the same component seams:
Directive/model/graph checks cover optional distinct seed lists, exact initial capacity, source locations, host-normalized labels, replacement type/rank mappings, stable palette fallback, accepted-status overrides, parts, external original URLs and hidden filtering before indexing/serialization.
Pure state cases assert complete-group expansion, both trace directions beyond three hops, diamonds, self-loops, isolated cycles, multiple roots, descendant restoration, page boundaries and independent page expansion sets. Rejected candidates retain the original state object.
Layout checks assert declared rank order, nonoverlapping finite rectangles, shared test placement, ports and routes that avoid node interiors. Browser cases inspect actual rendered boxes; measurements exercise real 200-node scenes rather than replacing ELK with a mock.
Detail and browser cases cover static host markup, local images, data images, srcset/CSS/media rewriting, remote fallback links, executable URL rejection, DOM ID/IDREF isolation, external-body fallback and static third-party widgets. Prototype-like IDs and script-closing text are data in classic registration scripts. Normal author prose is not treated as a hidden relationship field.
Browser cases cover two-line title clamping, complete wrapped IDs, empty and reviewed status borders, nonempty directional controls, missing-object text, whole-card original links, hover travel, bounded/scrollable panels, keyboard dismissal, native focus panning, pointer drag/wheel zoom and independent views.
Controlled layout completions exercise initial failure/retry, late-result discard, subsequent failure without commit, and focus ownership. Missing detail assets show an error and recover after the asset is restored.
Host/build cases exercise content hashes across incremental/no-op rebuilds, hidden-object removal, finalization failure and preservation of a prior fatal exception. Packaging cases inspect included assets/notices and the installed artifact, not a source-only import.
The behavioral transition/source/test mapping and editable published diagram are in Behavioral state machines. No later occurrence mode or selector is shipped.
Integration decisions within the accepted contracts
Sphinx-Needs 8.2.0 unconditionally registers unused GitHub service types during environment preparation. STV captures the author’s type schema before that injection, then includes any additional type actually used by host objects. Complete mapping validation therefore covers the authored schema and actual service data, without demanding mappings for unused injected service types or mutating the host schema. Host-normalized relationship defaults remain authoritative. This resolves the default five-type preset against the real host seam without changing the accepted architecture or adding another model.
Graph/detail publication uses content identities, null-prototype JSON data dictionaries and safe classic scripts. Runtime/CSS URLs also carry their asset digest. Current manifests reference only current fragments; obsolete output files can remain on disk after incremental builds. Distribute a clean output tree when changing visibility; STV cannot revoke older published copies.
Operating limits and deferred decisions
Windows Firefox’s local file loader needs short physical output paths. A detail filename contains two full SHA-256 digests; keep the complete physical asset path below the loader’s Windows path-length limit. File workflows use short fixture roots, including the large-project copy.
Supported release verification uses installed stable Firefox on Windows in headless automation. Touch, other browsers/platforms and occurrence mode are outside the initial scope.
Remote embedded resources are not fetched or cached. Details provide static unavailable-resource indications and explicit links; full widget interaction remains available on the original documentation page.
Performance measurements have no owner-agreed numeric latency/memory budgets yet. The hard 200-node limit is independently asserted. Recorded browser memory is the sum of process-tree working sets and can count shared pages more than once; it is not a private-memory or peak-allocation measurement.
The project license remains an owner decision before external distribution. Internal implementation/testing does not assign one. Dependency notices and the selected unmodified ELK bundle/source reference are included in the wheel.
Record integrity
Intermediate release attempts remain available even when their executable cases passed: source/test changes during review invalidate their frozen-input claim. Their source-stability outcomes are preserved. Only the final stable candidate can support the delivered verification disposition. The first report’s empty infrastructure Need label has a separately identified publication erratum; its original report and XML/JSON have not been overwritten.
v0.2.0 criterion assessment
The owner accepted the centered-layout/seed-selection concept in
Concept review round 6: accepted centered-layout and seed-selection amendment and invoked release-test against pre-release
revision bee5a15fed8e0b89a21b9cced09d7a1e67bb3c40.
The original assessment above remains historical. The delivered scope now covers
all ten system and 53 initial-release software requirements. The four
occurrence-mode requirements remain explicitly deferred and not_assessed.
The amendment’s criterion review adds the following observable evidence to the existing system/software assessment:
Initialize collapsed state (STV_SWR_003), Provide the artificial proj... (STV_SWR_051) and Page stakeholder entry links (STV_SWR_052): real seeded/seedless builds, initial project plus at most five entries, complete 201-entry/41-page traversal, empty projects, exact capacity counting, independent page memory and hidden-object filtering.
Separate transient hover an... (STV_SWR_050), Prefer direct facing-edge r... (STV_SWR_053), Set enlarged standard viewport (STV_SWR_054) and Order controls using target... (STV_SWR_055): measured hover versus explicit activation, real Escape outside view focus, cancellation, facing ports and safe obstacle/self-loop routes, computed 975px viewports, and stable semantic control ordering with replacement mappings. File/HTTP workflows exercise real pointer and keyboard behavior as well as the pure state/geometry checks.
Center parent boxes with co... (STV_SWR_056): exact five-child and project centering, shared-parent collision avoidance, retained identity, independent noncompeting parents, deterministic cycles/isolated nodes/variable heights, test lanes and final routes. The release review found that averaging collision displacement across the whole band moved an unrelated parent by 109.33px. Local least-displacement pooling fixes that defect within the accepted layout contract; the new regression would fail the previous implementation. Component boundaries, state transitions and ADR decisions are unaffected.
Confirm local interactive s... (STV_SWR_057): ID/title matching and result bounds, pointer and keyboard confirmation, typing/Escape, exact trimmed IDs, invalid/hidden/case-mismatched IDs, empty catalogs, repeated replacement, overview reset, independent seeded and seedless views, retained focus, and failed/stale layout preservation.
docs/test-scripts/test_release_seeds.pyextends the MVP smoke cases with these real file/HTTP paths; Node checks cover pure catalog and transition invariants.
A test synchronization defect was also corrected: the original-object workflow asserted a newly opened tab’s URL while Firefox still reported about:blank. The test now waits for the expected URL and actual Need element, preserving and strengthening the original criterion. The failed release attempt and separate diagnostics remain retained; isolated passes did not substitute for full qualification.
Final outcomes and artifact identities are in v0.2.0 final release validation. Canonical verification is promoted only against the full passing candidate and this criterion review; acceptance remains independent. Browser/platform, occurrence-mode, widget-interaction, numeric performance-budget and external-distribution licensing limits above still apply.
CR014 development criterion review
CR014 persistent details: development verification records the implemented persistent collection and author capacity setting against round 12. The new software and system cases pass normal, boundary, negative and asynchronous recovery paths. Hover timing/cancellation and safe URL/content transport regressions pass; explicit-Escape procedures now assert persistent preservation and use Close. Seed/page replacement, keyboard focus, narrow/wide docking and current-width Reset pass with the default ten and overridden three-box collections. Shared loading tests cover one-slot eviction, out-of-order completion, retry and clear. The repeated capacity, clean-wheel and source-archive checks pass. These results support the changed requirements’ verification fields, without changing unrelated partial/deferred verification or constituting release approval.
PR015 detail acceptance correction review
The owner’s findings are resolved within existing INSPECT/RENDER boundaries; PR015 detail acceptance corrections: development verification retains the actual failing baseline, 127 distinct passing selected checks and refined icon/ID repeat. The 720 px dock and 1120 px responsive boundary preserve usable graph space. Metadata defaults folded; pointer, Enter and Space update only its copied table with consistent icons and ARIA state. New boxes reset disclosure, reuse retains it. The artifact ID opens its real source Need; local body fragments retain their namespace. The actual normal demo and metadata state diagram/HTML embedding were visually reviewed. No arbitrary host/widget script is initialized and adjacent safe transport/offline/race/hover/package checks pass. Owner acceptance retest is pending; no release qualification or new capacity result is claimed.
CR015-018 development implementation, 2026-10-25
Input: accepted concept commit 2c4236a on concept-cr015-018, including
round 15 and the four owner-approved stakeholder amendments. Implementation is
an uncommitted development worktree, not a new release or deployed publication.
No dependency/version change is introduced. Development distributions remain
version 1.0.0 in an isolated directory; existing release assets are not replaced.
Delivered behavior
Collapse all relationships ... (STV_SHR_058) / Reset view-local exploratio... (STV_SWR_066): atomic Collapse all nodes keeps current roots, clears all page restoration caches, fits a fresh committed scene and preserves persistent cards. Layout retry repeats the rejected command.
Save the currently visible ... (STV_SHR_059) / Deliver a standalone visibl... (STV_SYS_011) / Capture a clipped inert vis... (STV_SWR_070), Save and release SVG downlo... (STV_SWR_071), Bundle safe export resource... (STV_SWR_072): inert visible-area SVG capture, local resource/font records, preserved crop/background/scroll layout and actual browser downloads. Errors retain the live view. Imported active SVG and outside-output paths fail; unreferenced images do not enter resource records.
Configure smaller text in b... (STV_SHR_060) / Validate and serialize inde... (STV_SWR_067), Scale copied detail typogra... (STV_SWR_068): independent author font scales default to 0.7, including loading/error wrappers and later inserted content. Baseline restoration and two-phase measurement avoid nested or repeated scaling, with no duplicated accessible measurement card.
Expand or collapse all Need... (STV_SHR_061) / Control rendered page metad... (STV_SWR_069): page-wide metadata controls on rootless and graph pages, preserving body/title, copied details and native individual operation. Unsupported encodings or absent host handlers produce disabled explanatory controls rather than partial success.
Implementation source and transition ownership are in Software elements and Behavioral state machines. Author settings and browser save-location guidance are in Getting Started. The interface changes bind the accepted design into actual function signatures and equivalent synchronous pre-paint measurement; no stakeholder behavior, component boundary or accepted ADR choice was weakened.
Executable coverage and actual developer checks
Canonical cases |
Executable selection |
|---|---|
Reset state, page caches an... (STV_TC_SW_051), Reset exploration without l... (STV_TC_SY_066) |
|
Validate font configuration... (STV_TC_SW_052), Configure and read both det... (STV_TC_SY_068) |
|
Snapshot XML, clip geometry... (STV_TC_SW_053), Embed only safe output-loca... (STV_TC_SW_055), Export and save the visible... (STV_TC_SY_067) |
|
Scope bulk metadata command... (STV_TC_SW_054), Bulk-fold host attributes w... (STV_TC_SY_069) |
|
Actual check sets are separate developer runs, not a manufactured aggregate release execution:
Complete frontend selection: 42 passed.
Complete non-browser selection excluding the separate capacity benchmark: 87 passed. Its package check uses the development wheel.
Broad Firefox selection excluding the separate capacity benchmark: 57 passed before the final export/page edge refinements.
Later delivered-scope/persistent-detail selection: 30 passed.
Final SVG/reset/font feature selection after crop/scroll corrections: 9 passed; page-only refinements are rechecked separately on the final build.
Clean offline wheel host without Node: passed, with required runtime/legal assets and no component tests in the runtime wheel. Final package hashes and source/test identity are retained in ignored developer build output.
Commands use .venv-implementation/Scripts/python.exe with installed Firefox
158.0.1 and geckodriver 0.37.1 for browser checks. For example:
node --test src/*/tests/*.test.js
.venv-implementation/Scripts/python.exe -m pytest -m "not browser" --ignore=docs/test-scripts/test_capacity.py -q
.venv-implementation/Scripts/python.exe -m pytest docs/test-scripts/test_view_controls.py -q
.venv-implementation/Scripts/python.exe scripts/check-coverage.py --strict
Run the Node command from frontend. Set STV_GECKODRIVER to the installed driver
and STV_WHEEL to the new wheel for consumer checks. Browser fixture downloads are
confined to automatically cleaned temporary folders. Early identified fixture
SVGs were moved out of Downloads into ignored build output.
Developer logs/JUnit, packages, source/test identity and visual examples are under
docs/_build/cr015-018*. They are transient local evidence, not committed release
records. The real generated demo and an exported standalone SVG were inspected
in Firefox. Visual checks exposed inherited-background and off-screen flow-layout
issues; the implementation now captures a backdrop and empty flow placeholders,
and browser checks assert visible reopened content rather than text presence only.
Verification boundary and handover
The four CRs are implemented. New requirement verification is conservatively
partial: these development checks do not constitute execution of every full
canonical case procedure, native Save As dialog qualification, the separate
30,000/90,000 capacity benchmark or a release qualification. The nine canonical
case designs remain accepted/not_run until their complete procedures are recorded;
no test_execution or historical release result was fabricated or overwritten.
Existing passing release claims retain their original candidate identities.
The development wheel/source archive are in
docs/_build/cr015-018/distributions. To replace the same version in a local
consumer, use pip with --force-reinstall --no-deps and the explicit wheel path
in that consumer’s environment. Readers need no Node or runtime service.
User inspection and separately authorized release-test work remain before any
new release/publication claim. Implementation changes have not been committed,
pushed or deployed by this invocation.
Final local verification identity
Source/test SHA-256: 0d79572952f63638efd7b433a6515370af15865cae2f2b7b38d65e2250fe0a50.
The complete input file hashes and archive hashes are in
docs/_build/cr015-018/source-identity.json. The wheel’s packaged stv.js was
compared byte-for-byte with the final rebuilt asset; runtime tests are excluded.
Final page-control retest: four passes, including explicit missing-host and body-target refusal and collision-safe ARIA row references. Final export retest: four passes (file/HTTP, one/ten cards), including actual temporary-folder file bytes, visible reopened content, and plaintext-password exclusion. Final resource and font validation selection: fourteen passes, including UTF-8/UTF-16 entity rejection. These supersede only their affected earlier developer checks; they are not combined into a fictitious release run.
Source whitespace checks exclude generated stv.js, whose vendored raw string literals legitimately retain trailing spaces; the generated bytes are not trimmed. Strict documentation, graph/evidence/test coverage and published diagram checks are recorded in Project maturity. No additional implementation commit was made.
CR019 remembered views
STATE snapshot/storage validation, RENDER checkbox/mount/D3 wiring, LAYOUT position/band restoration and INSPECT focus-free ordered reopening implement Store opt-in snapshots inde... (STV_SWR_073), Validate and reconstruct re... (STV_SWR_074), Restore presentation withou... (STV_SWR_075). CR019 development verification retains the exact source/test identities and development evidence. Current maturity and qualification limits are centralized in Project maturity; concept acceptance is in CR019 scoped concept review, round 16.