the map's second follow-up rulings: the digest has one shape, a frame holds what it is about and its roots are its goals, a registered result bears on an item rather than answering it, and an unconfirmed want weighs nothing (G1–G4)

Maintainer rulings, taken by one maintainer in conversation; the second maintainer's confirmation is outstanding (landed as the first commit of the map-as-canonical round's second follow-up, prompts/2026-10-22-map-as-canonical/PROMPT-followup-2.md; the confirmation is appended to the round's queue entry above rather than queued a third time). Two reviews of the branch at 4dc8811 agreed on the architecture and named the things to fix before merge; these four rulings answer them.

G1. The digest has one shape. derived.json's digest carries the same keys whatever previous_derivation returns: compared_with (a name, or none), statuses_flipped, root_statuses_flipped, levels_moved, events_new, revisions, wants_new, proposals_awaiting and whatever else the comparing branch emits — the lists empty and the counts zero when there is nothing to compare with. The DIGEST line says no previous derivation in that case, as now. A test that asserts a status flip supplies the previous derivation it needs explicitly — by writing the fixture's first derivation to the path previous_derivation reads, or by a test-only parameter — and never depends on a file a prior run left behind or on a branch existing. No test is loosened to tolerate a missing key.

G2. A frame holds what it is about; its roots are its goals. A frame gains a field holds (provisional name): a list of node ids the frame is about that are not goals of the ranking. A node in holds is attached (rule 1's "attached to nothing" is satisfied by membership), carries no role or preset, is no root, enters no status, value, stake or unit of the frame, and is listed in derived.json, the index (in_frames gains a membership: holds entry beside role) and the view under the frame as held. A node may be both held by one frame and a root of another. Then: every docket or is dissolved — its items go into holds of the frame it hung under, the docket node files are deleted (not superseded: they were placeholders the follow-up created, not claims anyone filed) — and every formal result that is a shared-claim root of a frame solely because the round filed it there (the 2026-10-22 and 2026-10-23 filings; the report says how it told these from roots that predate the round) moves from roots to holds. The same for the roots F4 moved to the legitimacy line frame on 2026-10-23 "because they bear on none of the three programmes": they are held, not goals, until a maintainer places them. The report gives root counts per frame before and after and the count of nodes held.

G3. Bearing, not answering. The render line becomes Registered results bearing on this item: and the index key answered_by becomes bearing (schema graph-index/3; §6.1 of graph/MAP-AS-CANONICAL.md updated). An item's item_status is unchanged by any registered result; closing is a maintainer act with a statement of record, as the regulations say.

G4. An unconfirmed want weighs nothing. A want with filed_from: quote and no priority event by its owner carries weight 0 in the ranking (it is listed, flagged owner has not confirmed, and the pooled list still asks for the word, as now); an explicit priority word gives it the word's weight; none still disowns. A want filed by its owner in a session (no filed_from) with no priority word keeps the Default medium, since the owner filed it. PRIORITY_WORDS and the want-weight function in checkers/graph.py, PROTOCOL.md §2 (owned wants) and the glossary say so; a design-log line records the Default change; the self-test gains the null input (a quoted want with no word weighs zero; with a word, the word).

Ruling G2 closes the queued entry How a docket is represented (ruling F6, 2026-10-23): a docket is no node; the work orders it held are held by the frame it hung under. The ruling text is the maintainer's and is not reworded.