WorldScript Studio investigated Chromium Embedded Framework (CEF) as a candidate next-generation desktop runtime between 2026-08-18 and 2026-08-20. That program is retired β CEF is not part of the target architecture. See ADR-0021, which supersedes ADR-0019 (CEF as next-gen runtime) and ADR-0020 (thin C++ host + Rust core binding choice).
Status of this document: Wave 0 is complete. ADR-0021 and the strategic-reset decision merged as
PR A (#406); the removal described below merged as PR B (#407).
A real Wave 2 spike got the Linux CEF sandbox genuinely operational for renderer processes, but
surfaced a structural blocker tracked as R-19 / Issue #405: under Linux's default Yama
ptrace_scope=1, Crashpad's PtraceStrategyDecider cannot produce sandboxed-renderer crash dumps
without weakening the sandbox. That trade was not acceptable. ADR-0021 has the full decision record,
including the alternatives considered and why each was rejected.
apps/desktop-cef/ (the C++ host + FFI-boundary Rust scaffold), scripts/cef/ (SDK-fetch/build/
proof tooling), .github/workflows/cef-learning-harness.yml (advisory-only CI, never part of the
required ci-success gate), and 14 of the 15 files under docs/cef/ (roadmap, risk register,
competency matrix, ownership manifest, binding scorecard, and knowledge-base articles).
docs/cef/UI-DOMAIN-STATE-CLASSIFICATION.md was not deleted β it was relocated to
docs/native/UI-DOMAIN-STATE-CLASSIFICATION.md
because its Redux domain/UI-state classification is independent of renderer and remains directly
useful for the future Rust Core migration-priority work.
Git history is the archive, not this directory. Once PR B merges, git log --oneline -- docs/cef/ apps/desktop-cef/ scripts/cef/ .github/workflows/cef-learning-harness.yml reconstructs the full
history regardless of any tag or branch. The companion PRs for this reset are the most direct
before/after reference: PR #384/#385 (Wave 1, DesktopPlatform boundary β still current, unaffected
by this retirement) and the Wave 0 strategy-reset (PR A) and cleanup (PR B) PRs that remove the paths
above.
Companion GitHub state: PR #404 (the real sandbox-enable attempt) is closed unmerged, not deleted β
its investigative work remains visible history. Issue #405 (R-19) is closed as "no longer applicable
to the target architecture," explicitly not "fixed" β the Crashpad root cause it documented remains
real and correctly diagnosed; it is simply no longer WorldScript's problem to solve. Both closed once
ADR-0021 existed on main (after PR A merged), independent of PR B's own merge status.