Waves Browser WBP-00/WBP-01 Baseline

Status: accepted baseline Captured: 2026-07-25 (evidence artifact timestamp is UTC) Owner lane: browser

Scope and Decisions

This baseline closes the adoption and shell-integration gates for the adopted Authentic Core, Modern Console direction. It is evidence for the current shell, not a visual golden and not a device-specific compatibility claim.

The initial Class C Reference profile uses:

• a neutral, non-vendor identity;
• 20 engine-owned logical columns, matching the existing WaveNav/browser default;
• no browser-authored fixed row count, font metric, softkey precedence, or hit-region model before the WBP-06 frame/input contract;
• a 1024 by 768 CSS-pixel default test viewport and an 880 by 640 CSS-pixel minimum test viewport, matching browser/src-tauri/tauri.conf.json;
• the existing Up, Select, and Down controls only as the current adapter. They do not imply future softkey slots or task precedence.

The primary launch audience assumption is WAP/WML implementers and emulator evaluators. Current evidence is the product’s transport-route inspection, correlation/timeline surfaces, canonical conformance examples, raw debug path, and executable story workflow. Preservationists and learners remain important secondary audiences; WBP-04 owns validation and onboarding for them.

Reference Environment

• MacBook Pro MacBookPro18,4
• Apple M1 Max, 10 logical cores
• 64 GB memory
• macOS 26.5.2 (Darwin 25.5.0), arm64
• Node.js 22.22.1
• headless Chromium 151.0.7922.34
• production-built browser-story.html using the real WaveNav WASM engine and deterministic local fixture fetch adapter

The checked-in machine-readable result is evidence/wbp-00-01/baseline.json. It contains every sample, environment details, window geometry, tab order, and keyboard activations.

Baseline Measurements

Twenty samples were captured for each metric:

Metric	Boundary	p50	p95
Startup	navigation start to initial local deck committed and rendered	53.8 ms	57.9 ms
Local navigation	local-example change to selected deck committed and rendered	7.3 ms	7.5 ms
Input/render	browser ArrowUp/ArrowDown event to engine focus plus focused DOM render	7.3 ms	7.5 ms

These numbers describe this reference machine and deterministic local path. They are comparison baselines, not release budgets and not evidence for native network or gateway latency.

Reproduce

From the repository root with the pinned Node/pnpm toolchain active:

cd engine-wasm/engine
wasm-pack build --target web --out-dir ../pkg
cd ../..
WAVES_BASELINE_RUNS=20 pnpm test:baseline:waves
pnpm test:story:waves

Set WAVES_BASELINE_OUTPUT_DIR to preserve a particular run outside the default ignored engine-wasm/host-sample/test-results/waves-baseline directory. Relative output paths resolve from the repository root.

Window and Keyboard Evidence

Both automated window runs assert:

• no horizontal overflow or clipped shell landmark;
• the handset input controls are inside the initial viewport;
• the same shell/controller/WASM path reaches a committed local deck;
• every enabled native control and disclosure is reachable in a stable 28-control tab sequence (29 after Back enables);
• Utility Rail, Welcome/Help, Local Example Notes, Developer Tools, and Raw WML disclosures open with Enter;
• Back enters the first toolbar tab stop when engine history enables it;
• Back, Reload, and Health activate from the keyboard.

At 1024 by 768 the utility rail starts open. At 880 by 640 it starts collapsed, the handset stage uses the full available width, and Up/Select/Down remain visible without initial vertical scrolling.

Default 1024 by 768 shell baseline screenshot

Minimum 880 by 640 shell baseline screenshot

The screenshots are historical baseline evidence only. WBP-02 must not treat them as target goldens.

The unavoidable manual gap is native macOS decoration/menu appearance and platform-drawn focus styling inside the packaged Tauri window. The webview layout, disclosure behavior, input routing, and configured window dimensions are automated; native menu wiring remains covered by Tauri Rust checks.

Closure Verification

The accepted run completed:

• browser frontend: 181 unit tests, ESLint, TypeScript typecheck, Prettier check, and production Vite build;
• executable behavior: all 9 Waves-targeted stories;
• baseline: 20 runs per metric plus default/minimum layout and keyboard audits;
• contracts: Rust/AST regeneration with zero drift;
• Tauri: Rust formatting, Clippy with warnings denied, and 56 non-ignored tests across the host library and contract generator (the five external-Kannel smokes remain intentionally ignored);
• documentation: changed-document local-link check, worklist drift, requirement/status drift, Astro check, Astro production build, and Git whitespace check.

The first browser keyboard pass directly exposed global Enter interception on native controls. The routing fix now leaves buttons, inputs, selects, summaries, links, and editable host content to their native behavior while the focused handset viewport continues to route engine keys.

Phase 1 Integration Seams (Historical)

The #343 decomposition remains authoritative. WBP-02 through WBP-04 landed in #344, #346, and #347 while this integration gate was in progress; WBP-05 then landed in #356 before the final #348 integration merge. These rows preserve the maintenance seams used by the completed Phase 1 slices; they are no longer next-work pointers:

Slice	Safe leaf seam	Integration constraint
WBP-02	app/shell/handset-stage-template.ts plus scoped component styles/tokens	Do not change 20 logical columns or infer frame, softkey, or hit-region semantics. Root styles.css integration remains single-owner.
WBP-03	app/shell/navigation-toolbar-template.ts and browser-owned controller callbacks	Preserve ShellEventBindingActions, current navigation ordering, Local/Network behavior, and transport truthfulness.
WBP-04	new start/help leaf templates and distinct engine-wasm/examples/source/* story files	Root-shell insertion, waves-copy.ts, and generated example-manifest updates require the integration owner. Tutorial decks use the ordinary engine path.
WBP-05	landmark/ARIA/focus changes within existing leaf templates and scoped tests	Reuse native-control keyboard behavior. Do not create a DOM interpretation of WML or add future engine action semantics. The now-completed additive WBP-05A follow-up closed the single-announcement and rendered-evidence gap found at the Phase 1 checkpoint.

The root browser-shell-template.ts, global styles.css, waves-copy.ts, and generated example manifest are high-conflict integration surfaces. The phase slot remains the reserved browser-owned insertion point; the handset stage retains the current presenter adapter; the developer drawer remains a top-level sibling rather than utility-rail content.