ADR-0066: WITHDRAW — the Author Side May Reclaim a Version from Review¶
- Status: accepted
- Date: 2026-07-13
- Deciders: Nicolas Burri ("I found a black hole… it would be great if the submitting parties could reclaim the document"; W-1a/W-2a: "I agree with both suggestions"), during the extended test round.
- Amends: the fixed lifecycle (ADR-0003 lineage): IN_REVIEW gains a second exit besides REJECT/APPROVE-completion — WITHDRAW → DRAFT, initiated by the author side.
- Relates to: ADR-0064 (author group semantics + the frozen-bindings rule this completes), ADR-0030 (policy snapshot — WITHDRAW is deliberately NOT policy-configurable for now).
Decision¶
- New transition WITHDRAW (IN_REVIEW → DRAFT). Permitted for any holder of the document's AUTHOR workflow-role group (W-1a — accountability lives in the role group, not the individual submitter; no black hole when the submitter is unavailable). Not a policy knob; fixed rule (a knob can be added later without schema change).
- Allowed at any point while IN_REVIEW (W-2a) — including after partial approvals: finding a problem after one of two approvals is exactly when reclaiming matters. Consequences of a withdraw: the review round closes; already-given approvals are VOIDED (they endorsed content about to change); open APPROVAL_REQUIRED inbox tasks are cancelled; the frozen content returns to the draft area (the REJECT unfreeze machinery generalizes); the workflow-role bindings unlock again (ADR-0064 D-2a — withdraw-and-resubmit is the sanctioned way to change reviewers mid-flight, now actually possible).
- Audited as
WITHDRAWN(actor, optional comment), visible in the version history and the audit trail — a reviewer whose approval evaporated can see exactly why. - Resubmission starts a fresh round with a fresh pin (content + workflow-roles snapshot); four-eyes evaluates against the new round's submitter as always.
Related decision recorded the same day (no separate ADR)¶
Reviewers act by COMMENTS and verdicts, never by direct edits (Nicolas's question, resolved by the regulatory analysis): reviewer edits would break freeze integrity (the approved artifact must be the submitted artifact), collapse four-eyes (approving one's own fix), and blur authorship attribution (§4.2.3 designated-authority clarity). The loop is: comment/reject → author withdraws or receives rejection → fixes in DRAFT → resubmits. The comments feature is a future design arc ("suggested changes" as its possible evolution); this paragraph pins only the principle: content mutation stays with AUTHOR-side roles in DRAFT state.