Slate DOM v2 Bridge Proof Ralph Plan

Goal

Ship the first honest DOM-boundary proof for Slate v2:

• .tmp/slate-v2/packages/slate-dom-v2
• one narrow bridge over slate-v2 snapshots and runtime ids

The slice must prove that DOM lookup and boundary containment can work without path-only weak-map luck.

Locked Constraints

• no full React runtime rewrite
• no history
• no clipboard lane yet
• no IME gauntlet yet
• no browser-state logic inside slate-v2
• stop if DOM ownership still leaks between nested editors

Planned Proof Slice

Primary target:

• identity-backed DOM path lookup
• nested-editor root containment
• explicit rejection of points outside the editor boundary

Representative pressure:

• #5947
• #5938
• #4789

Execution Plan

Phase 1: Repo Grounding

Status: completed

• inspect current slate-dom seams and reusable patterns
• inspect current test seams for DOM utilities
• decide the smallest slate-dom-v2 package shape

Phase 2: Slice Definition

Status: completed

• freeze the minimal public surface
• freeze the first red-test lanes
• freeze the bridge ownership model for nested editors

Phase 3: Implement packages/slate-dom-v2

Status: completed

• scaffold package
• implement editor-root registration
• implement element-to-runtime-id registration
• implement runtime-id-backed findPath
• implement guarded toSlatePoint or equivalent boundary-safe point lookup
• add tests

Phase 4: Verification And Review

Status: completed

• targeted typecheck
• targeted tests
• root build or lint wiring if needed
• architect review
• deslop pass
• regression re-verification

Known Failure Conditions

• parent editor can still resolve child editor DOM
• path lookup still depends on stale path-only maps
• outside points still resolve into the wrong editor
• the bridge needs render-timed React state to stay coherent

Progress Log

2026-04-02

• created new Ralph context snapshot for the DOM bridge branch
• read the locked v2 DOM contract and current slate-dom entrypoints
• selected the first proof slice around runtime-id lookup and nested-editor containment
• explorer findings confirmed the right cut: start from containment plus a brand-new runtime-id registry, not the current DOMEditor tarpit
• implemented packages/slate-dom-v2 with:
  • editor root mounting
  • per-editor DOM node to runtime-id binding
  • runtime-id-backed findPath
  • guarded toSlatePoint
• added proof tests for:
  • nested editor containment
  • id-backed path lookup after reorder
  • outside-point rejection
• green so far:
  • fnm exec --using 22 yarn install
  • fnm exec --using 22 yarn tsc --project packages/slate-dom-v2/tsconfig.json --noEmit
  • fnm exec --using 22 yarn workspace slate-dom-v2 test
  • fnm exec --using 22 yarn build:rollup
  • fnm exec --using 22 yarn lint:eslint
  • fnm exec --using 22 yarn lint:prettier
  • lsp diagnostics directory for packages/slate-dom-v2: 0 errors
• waiting on architect verification before the deslop pass
• architect review first rejected the slice for three real blockers:
  • broken CommonJS consumption for the new proof packages
  • stale root ownership after remount
  • element-container DOM points not normalizing into text points
• those blockers were fixed and pinned with extra tests
• final architect verdict: APPROVED
• deslop pass was a no-op after the blocker fixes; there was no further honest simplification left in the changed files without weakening the proof
• post-deslop regression re-verification is green:
  • fnm exec --using 22 yarn install
  • fnm exec --using 22 yarn tsc --project packages/slate-dom-v2/tsconfig.json --noEmit
  • fnm exec --using 22 yarn workspace slate-dom-v2 test
  • fnm exec --using 22 yarn build:rollup
  • fnm exec --using 22 yarn lint:eslint
  • fnm exec --using 22 yarn lint:prettier
  • fnm exec --using 22 yarn node -e "require('slate-v2'); require('slate-dom-v2')"
  • lsp diagnostics directory for packages/slate-dom-v2: 0 errors