Slate v2 IME / Mobile / Browser File Ledger

Historical batch/review note. Live browser/input scenario-proof truth now lives in true-slate-rc-proof-ledger.md. Live legacy-file closure truth now lives in release-file-review-ledger.md. Keep this file for archaeology and batch history, not as the active proof authority.

Purpose

Historical batch/review record for the reopened IME/mobile/browser RC lane.

Use the same blunt rule as the other ledgers:

[x] reviewed
[ ] still needs deep review

Do not use this file as passive archaeology. Every behavior-bearing legacy file must map to one or more scenario rows.

Ground Rules

missing proof on a behavior-bearing row blocks by default
only engine-internal helpers with no user-visible browser/input behavior may be omitted
older “better-cut” language in release docs does not close this lane by itself

Bootstrap Seed

Pinned slate-v2 diff command result:

git -C /Users/zbeyens/git/slate-v2 diff --name-status --diff-filter=ACDMRTUXB origin/main...HEAD -- \
  packages/slate-react \
  packages/slate-dom \
  packages/slate-browser \
  playwright/integration/examples \
  site/examples/ts

Current result:

EMPTY

Read:

there are no still-unreviewed browser/input rows hiding in the current origin/main...HEAD diff for those target directories
the seed therefore comes from the live proof lanes plus the legacy behavior-bearing patch files themselves

Dual-Axis Behavior / Parity Ledger

Review	Scenario	Legacy file(s)	Current owner / equivalent	Proof lane	Browser scope	Expected outcome	Actual outcome	Artifact links	Owner	Status
[x]	placeholder IME	`packages/slate-react/src/components/editable.tsx`, `packages/slate-react/src/components/string.tsx`	`Editable`, `EditableText`, placeholder example	`placeholder-ime.test.ts`, `firefox-direct-ime.test.ts`, `webkit-direct-ime-ceiling.test.ts`, `ime-proxy.test.ts`, `proof:agent-browser:ios:placeholder-input:local`, `proof:appium:ios:placeholder-input:local`, `proof:appium:android:placeholder-input:local`	`automated-direct: chromium + firefox + Android Chrome emulator`, `automated-proxy: webkit-desktop`, plus `setup-green / behavior-red: iOS Simulator Safari`	empty placeholder composition behaves safely enough to support the strong RC claim	Chromium direct proof is green on the FEFF-backed placeholder path, Firefox now has a stronger direct composition lane via Playwright `keyboard.insertText`, desktop WebKit has an explicit direct-input ceiling probe showing that `insertText` commits cleanly but does not expose composition-specific input events there, Android direct typing proof is green, and direct Appium iOS Safari setup is green. Current iOS behavior proof remains red through Appium/XCUITest `value` (`blockTexts: ""`, repeated `input:undefined:[null]`, `slateSelection: 0.0:0	0.0:0`), and native-context coordinate tapping still fails to surface an iPhone keyboard. The current` agent-browser` iOS provider is also not trustworthy on these routes because it often exposes only the Next shell without the editor node	placeholder-ime.test.ts, firefox-direct-ime.test.ts, webkit-direct-ime-ceiling.test.ts, ime-proxy.test.ts, 2026-04-03-jsdom-contenteditable-composition-is-not-a-trustworthy-ime-proof.md, 2026-04-12-appium-ios-safari-loads-local-slate-routes-but-xcuitest-value-does-not-drive-contenteditable.md, package.json	`playwright + site/examples + browser-mobile transports`
[x]	no-FEFF placeholder IME	`packages/slate-react/src/components/editable.tsx`, `packages/slate-react/src/components/string.tsx`	`Editable`, `EditableText`, `placeholder-no-feff` example	`placeholder-ime.test.ts`, `firefox-direct-ime.test.ts`, `ime-proxy.test.ts`, `proof:agent-browser:ios:placeholder-no-feff-input:local`, `proof:appium:ios:placeholder-no-feff-input:local`, `proof:appium:android:placeholder-no-feff-input:local`	`automated-direct: chromium + firefox + Android Chrome emulator`, `automated-proxy: webkit-desktop`, plus `setup-green / behavior-red: iOS Simulator Safari`	no-FEFF empty placeholder path composes like the relied-on legacy path or is explicitly justified as a changed contract	Chromium composition proof is green on the line-break no-FEFF path, Chromium delayed per-key typing is now also green, Firefox has direct composition proof on the same surface via Playwright `keyboard.insertText`, desktop WebKit keeps the proxy lane, and Android direct no-FEFF typing proof is green. Direct Appium iOS Safari setup is green on the route, but typing remains red with the same XCUITest `value` no-op shape as FEFF (`blockTexts: ""`, repeated `input:undefined:[null]`, `slateSelection: 0.0:0	0.0:0`). The current` agent-browser` iOS provider is route-shell-broken here, so it should not be treated as behavior evidence	placeholder-ime.test.ts, firefox-direct-ime.test.ts, ime-proxy.test.ts, 2026-04-12-appium-ios-safari-loads-local-slate-routes-but-xcuitest-value-does-not-drive-contenteditable.md, package.json	`playwright + site/examples + browser-mobile transports`
[x]	inline-edge IME	`packages/slate-react/src/components/editable.tsx`, `packages/slate-react/src/components/string.tsx`	`Editable`, `EditableText`, inline-edge example	`inline-edge-ime.test.ts`, `firefox-direct-ime.test.ts`, `webkit-direct-ime-ceiling.test.ts`, `ime-proxy.test.ts`, `proof:agent-browser:ios:inline-edge-input:local`, `proof:appium:android:inline-edge-input:local`	`automated-direct: chromium + firefox + iOS Simulator Safari + Android Chrome emulator`, `automated-proxy: webkit-desktop`, plus `direct-input ceiling: webkit-desktop`	inline-edge composition lands on the intended text leaf with legacy-equivalent behavior	Chromium direct proof is green after semantic selection setup, Firefox now has a stronger direct composition lane via Playwright `keyboard.insertText`, desktop WebKit has an explicit direct-input ceiling probe showing that `insertText` commits cleanly but stays below composition-specific proof, and the packaged iOS Simulator/Appium edge-input lanes are green once the runner collapses DOM selection onto the leading zero-width text leaf before typing	inline-edge-ime.test.ts, firefox-direct-ime.test.ts, webkit-direct-ime-ceiling.test.ts, ime-proxy.test.ts, 2026-04-04-inline-edge-ime-proofs-should-set-selection-semantically-before-composition.md, package.json	`playwright + site/examples + browser-mobile transports`	`open`
[x]	void-edge IME	`packages/slate-react/src/components/editable.tsx`, `packages/slate-react/src/components/string.tsx`	`Editable`, `EditableText`, void-edge example	`void-edge-ime.test.ts`, `firefox-direct-ime.test.ts`, `webkit-direct-ime-ceiling.test.ts`, `ime-proxy.test.ts`, `proof:agent-browser:ios:void-edge-input:local`, `proof:appium:android:void-edge-input:local`	`automated-direct: chromium + firefox + iOS Simulator Safari + Android Chrome emulator`, `automated-proxy: webkit-desktop`, plus `direct-input ceiling: webkit-desktop`	void-like IME behavior matches the relied-on spacer semantics	Chromium direct proof is green on the real spacer structure, Firefox now has a stronger direct composition lane via Playwright `keyboard.insertText`, desktop WebKit keeps an explicit direct-input ceiling probe instead of fake direct composition proof, and the packaged iOS Simulator/Appium edge-input lanes are green with the same zero-width selection-collapse primitive used before typing	void-edge-ime.test.ts, firefox-direct-ime.test.ts, webkit-direct-ime-ceiling.test.ts, ime-proxy.test.ts, 2026-04-04-void-like-zero-width-ime-proofs-need-the-real-void-spacer-structure.md, package.json	`playwright + site/examples + browser-mobile transports`	`open`
[x]	blur/focus selection recovery	`packages/slate-react/src/components/editable.tsx`, `packages/slate-react/test/react-editor.spec.tsx`	`Editable`, `ReactEditor`, `rich-inline` reset/refocus flow	`surface-contract.tsx`, `rich-inline.test.ts`, `test:slate-browser:desktop-parity:local`	`headless + automated-direct: chromium, firefox, webkit-desktop`	focus restore behaves at least as safely as legacy across the required browser matrix	current focus init and mid-transform safety are green in headless proof, and the rich-inline blur/focus reset row is now green on Chromium, Firefox, and desktop WebKit; real iOS Safari remains unproved	surface-contract.tsx, rich-inline.test.ts, package.json	`packages/slate-react + playwright`	`open`
[x]	transient DOM-point gap / mutation-repair / fail-closed focus restore	`packages/slate-react/src/components/restore-dom/restore-dom.tsx`, `packages/slate-react/src/components/restore-dom/restore-dom-manager.ts`, `packages/slate-react/src/components/editable.tsx`	mounted `Editable`, `ReactEditor`, `DOMBridge`	`surface-contract.tsx`, main IME/browser rows, structural split/join row	`headless`, `automated-direct: chromium, firefox, android emulator`, plus `automated-proxy: webkit-desktop`	transient DOM bridge gaps fail closed without hidden runtime errors and without regressing user-visible behavior	current focus-time transient DOM-point failure is covered and fails closed; main IME rows are green without `restore-dom`; structural Enter/Backspace churn is now green on Chromium and Android through editor-owned keydown paths; the deleted restore-dom family is now treated as a rerender-era guard that is no longer a standalone current blocker	surface-contract.tsx, 2026-04-09-slate-react-focus-restore-must-fail-closed-on-transient-dom-point-gaps.md, 2026-04-12-restore-dom-was-a-rerender-era-guard-not-a-current-v2-runtime-need.md, 2026-04-12-structured-enter-and-backspace-need-editor-owned-keydown-paths.md	`packages/slate-react + packages/slate-dom`	`closed`
[x]	zero-width selection normalization	`packages/slate-react/src/components/string.tsx`, `packages/slate-react/src/components/editable.tsx`	`DOMBridge`, `slate-browser` selection helpers, zero-width matrix example	`zero-width-matrix.test.ts`, `bridge.ts` tests, `test:slate-browser:desktop-parity:local`	`automated-direct: chromium, firefox, webkit-desktop` plus headless bridge proof	DOM/Slate round-trip around zero-width leaves stays normalized instead of leaking sentinel offsets	current bridge proof is green, and the zero-width matrix row is now green on Chromium, Firefox, and desktop WebKit; real iOS Safari remains unproved	zero-width-matrix.test.ts, 2026-04-03-zero-width-dom-selection-bridges-must-normalize-both-directions.md, 2026-04-04-slate-browser-playwright-helpers-must-normalize-zero-width-selection-and-wait-for-selection-sync.md, package.json	`packages/slate-dom + slate-browser`	`open`
[x]	post-composition undo / redo	`packages/slate-react/src/components/editable.tsx`	`withHistory(createEditor())`, browser/history proof stack, IME proof surfaces	`placeholder-ime.test.ts`, `inline-edge-ime.test.ts`, `void-edge-ime.test.ts`, `test:slate-browser:ime:local`	`automated-direct: chromium`	undo and redo immediately after composition are directly proved on the placeholder/inline-edge/void-edge rows	dedicated IME history proof now exists on the FEFF placeholder path, no-FEFF placeholder path, inline-edge path, and void-edge path in Chromium; broader platform/browser IME parity is still blocked elsewhere	placeholder-ime.test.ts, inline-edge-ime.test.ts, void-edge-ime.test.ts, package.json	`packages/slate-history + playwright`	`open`
[x]	Android composition / diff / flush	`packages/slate-react/src/hooks/android-input-manager/android-input-manager.ts`, `packages/slate-react/src/hooks/android-input-manager/use-android-input-manager.ts`, `packages/slate-react/src/components/editable.tsx`	missing direct equivalent; current owner hypothesis is mounted `Editable` plus browser proof lanes	`android-tests.test.ts`, `test:slate-browser:android-proxy:local`, `proof:appium:android:local`, `proof:appium:android:placeholder-input:local`, `proof:appium:android:inline-edge-input:local`, `proof:appium:android:void-edge-input:local`, `proof:appium:android:split-join:local`, `proof:appium:android:empty-rebuild:local`, `proof:appium:android:remove-range:local`, `android-split-join.test.ts`, `android-empty-rebuild.test.ts`, `android-remove-range.test.ts`, `android-special-structural.test.ts`	`automated-proxy: mobile Chromium`, `automated-direct: Android Chrome emulator IME rows plus split/join, empty/delete-rebuild, remove-range, and special structural rows`, plus `remaining-android-specific: autocorrect / glide / voice`	Android-specific composition/diff/flush behavior is directly proved or explicitly justified as engine-internal	the Android hub exists, the mobile Playwright proxy lane is green on placeholder/inline-edge/void-edge plus IME undo/redo, Appium now has direct green packaged rows for placeholder, inline-edge, void-edge, split/join, empty/delete-rebuild, and remove-range on Android Chrome emulator, and the matching Chromium rows for split/join, empty, remove, and the structural `special` subcases are also green. The remaining Android-specific open slice is now the keyboard-feature lane only. Direct local probes can show `keyboardShown: true` and switch to `NATIVE_APP`, but expose zero Gboard candidate nodes, and hardware keycodes only yield literal `cant` insertion, so autocorrect / glide / voice are explicitly tooling-blocked on the current local stack	android-tests.test.ts, android-split-join.test.ts, android-empty-rebuild.test.ts, android-remove-range.test.ts, android-special-structural.test.ts, package.json, 2026-04-11-appium-android-setup-proof.md, 2026-04-12-structured-enter-and-backspace-need-editor-owned-keydown-paths.md, 2026-04-12-structural-break-proof-rows-need-selection-sync-before-follow-up-typing.md, 2026-04-12-appium-android-chrome-can-show-keyboard-state-without-exposing-gboard-candidates.md, packages/slate-react/src/hooks/android-input-manager/android-input-manager.ts	`platform parity lane`	`tooling-blocked / external`
[x]	iOS Safari / WebKit composition / focus	`packages/slate-react/src/components/editable.tsx`	`Editable`, `ReactEditor`, `DOMBridge`	desktop parity lane, explicit WebKit direct-input ceiling lane, composition proxy lane, `proof:agent-browser:ios:local`, `proof:agent-browser:ios:placeholder-input:local`, `proof:agent-browser:ios:inline-edge-input:local`, `proof:agent-browser:ios:void-edge-input:local`, plus missing real-device lane	`automated-direct: webkit-desktop` for focus/selection, `direct-input ceiling: webkit-desktop`, `automated-proxy: webkit-desktop` for composition, `automated-direct: mixed iOS Simulator Safari rows`, `tooling-blocked: broader iOS Safari composition/focus`	iOS Safari / WebKit composition and focus behavior is directly proved or explicitly justified as non-behavioral	desktop WebKit has direct proof for zero-width normalization and blur/focus recovery, an explicit direct-input ceiling probe showing that Playwright `insertText` does not expose composition-specific input events there, and a green browser-level proxy composition lane on the IME surfaces. On iOS Simulator Safari, direct Appium route/setup is green, inline-edge and void-edge have direct green packaged rows, but placeholder and no-FEFF placeholder typing remain behavior-red through XCUITest `value`. Broader iOS Safari composition/focus remains tooling-blocked, and the external plan now lives in 2026-04-12-ios-safari-broader-composition-focus-external-evidence-plan.md	rich-inline.test.ts, zero-width-matrix.test.ts, webkit-direct-ime-ceiling.test.ts, ime-proxy.test.ts, 2026-04-11-slate-browser-agent-browser-ios-setup-proof.md, 2026-04-12-appium-ios-safari-loads-local-slate-routes-but-xcuitest-value-does-not-drive-contenteditable.md, package.json	`platform parity lane`	`tooling-blocked / non-RC`
[x]	Firefox composition / selection recovery	`packages/slate-react/src/components/editable.tsx`	`Editable`, `EditableBlocks`, `ReactEditor`, `DOMBridge`, drag/drop cleanup example, table multi-range example, nested-editable focus example	desktop parity lane plus direct composition lane plus proxy backstop plus dedicated drag/drop, table multi-range, and nested-editable lanes	`automated-direct: firefox` for selection recovery, IME composition, dragged-node-unmount cleanup, table multi-range preservation, and nested-editable focus bounce, `automated-proxy: firefox + webkit-desktop` backstop	Firefox composition, selection recovery, dragged-node-unmount cleanup, table multi-range preservation, and nested-editable focus bounce are directly proved or explicitly justified as non-behavioral	Firefox now has direct browser proof for blur/focus selection recovery, zero-width normalization, direct composition on placeholder, no-FEFF placeholder, inline-edge, and void-edge via Playwright `keyboard.insertText`, plus dedicated drag/drop, multi-range table, and nested-editable focus lanes. Local Firefox/browser parity is exhausted; the remaining work is external Android keyboard-feature evidence and broader iOS evidence, not another Firefox hole	rich-inline.test.ts, zero-width-matrix.test.ts, firefox-direct-ime.test.ts, drag-drop-cleanup.test.ts, table-multi-range-firefox.test.ts, firefox-nested-editable-focus.test.ts, ime-proxy.test.ts, package.json, 2026-04-12-firefox-drag-drop-proof-needs-example-owned-drop-mutation-and-document-level-drag-cleanup.md, 2026-04-12-firefox-table-multi-range-proof-needs-native-table-selection-and-a-multi-range-sync-guard.md, 2026-04-12-firefox-nested-editable-focus-proof-needs-a-real-nested-contenteditable-surface.md	`platform parity lane`	`local-closed`

Legacy Slate React High-Risk Files

packages/slate-react/src/components/editable.tsx Note: old browser/input complexity center, including Android manager wiring, restore-dom, Firefox/Safari compat, and history/browser event policy.
packages/slate-react/src/components/string.tsx Note: zero-width, void spacer, and placeholder DOM semantics live here.
packages/slate-react/src/hooks/android-input-manager/android-input-manager.ts Note: explicit Android composition/diff/flush subsystem.
packages/slate-react/src/hooks/android-input-manager/use-android-input-manager.ts Note: Android-only hook integration seam.
packages/slate-react/src/components/restore-dom/restore-dom.tsx Note: mutation-observer restore seam used on Android and weird DOM drift.
packages/slate-react/src/components/restore-dom/restore-dom-manager.ts Note: buffered mutation restore logic during composition-sensitive phases.
packages/slate-react/test/react-editor.spec.tsx Note: direct legacy focus/selection recovery proof rows.

Current Slate v2 Runtime Files

packages/slate-react/src/components/editable.tsx Note: current mounted bridge, selection sync, DOM commit, focus recovery.
packages/slate-react/src/plugin/react-editor.ts Note: reviewed; current owner of focus/blur, target detection, window/root lookup, and DOM selection restore. It does not carry legacy Android/Safari/ Firefox composition-specialist machinery.
packages/slate-dom/src/bridge.ts Note: reviewed; current owner of DOM point/range translation and zero-width normalization. It carries the browser boundary seam, but not platform-specific IME orchestration.
packages/slate-react/test/surface-contract.tsx Note: current direct focus/runtime owner file for the mounted bridge.
packages/slate-react/test/runtime.tsx Note: broad runtime contract owner, but not yet the IME/mobile/browser parity authority by itself.

Current Browser Proof Files

Current Docs / Risk Framing Files

Current Read

Chromium IME/browser proof is real and worth keeping
the strong zero-regression RC claim is still blocked by platform/browser proof holes
the remaining behavior-bearing rows are now:
- Android keyboard-feature behavior beyond ordinary typing and structural rows
- broader iOS Safari / WebKit composition/focus beyond the currently green packaged lanes
the local Firefox/browser lane is exhausted
all three now have a cleaner manual-device path because the IME surfaces emit debug JSON and HTML dumps through ?debug=1
ReactEditor and DOMBridge are now reviewed as current owners
the remaining red is external/tooling-limited platform proof, not unread core owner files
row-specific manual capture scaffolds now exist under:
- android-keyboard-features
- ios-safari-composition-focus

Exact Legacy Compat Behaviors Still Unproved

Android

beforeinput is non-cancelable on Android, so legacy routed it through the Android input manager instead of the normal DOM path
Android IMEs can re-apply their own selection after Slate sets one, which is why legacy forced selection twice across animation-frame and timeout phases
Gboard spellchecker state needed an extra forced selection change after the animation-frame selection update
the current local Appium + Chrome stack can prove keyboard visibility but not expose or accept Gboard candidates, so autocorrect / glide / voice remain external-evidence rows

Primary legacy evidence:

Safari / WebKit

Shadow-root selectionchange in WebKit needed the odd execCommand('indent') / deselect workaround path
Safari can dispatch beforeinput(insertFromComposition) before compositionend, so composition state had to be cleared early
Safari blur could leave selection ranges behind even after the editable lost focus
Safari paste InputEvents could miss the Slate fragment payload, requiring the clipboard fallback path

Primary legacy evidence:

packages/slate-react/src/components/editable.tsx

Firefox

Firefox selection reads needed special handling because the “normal selection way” could lie
nested editable focus had to be bounced back to the main editor to avoid keyboard-navigation breakage
Firefox multi-range table selection could be broken by calling setDomSelection
Firefox drop flows needed global dragend listening because dragged nodes could unmount before their own dragend

Primary legacy evidence:

packages/slate-react/src/components/editable.tsx