Brick 24: common-line cutting — the shared-edge web design, engine + Nest-tab disclosure block

kerfmaster pin list · The interface — recorded intent (build later)

recordStates what the code does today. Auditable against the code — this is the only type that can go stale.

Content last changed 2026-09-06 — computed from the item itself, not typed.

Built 2026-09-01 (Aristide 'go'). The pin common-line-cutting recorded: 'how our safety-order rules map onto a shared-edge web is design work not yet done.' common_line.py does that design and records it in its own docstring:

• The welded unit is the BAND — parts joined along coincident straight edges. The default band is a COLUMN (horizontal shared edges, parts stacked in Y) because fill-Y-first is the ruled fill pattern; vertical shared edges build row bands; 'both' welds a rectangle grid into one block.
• min_gap keeps governing everywhere an edge is not deliberately shared — between bands and at the sheet edge. Gap zero applies only ON the shared edges (the ruled deliberate exception, recorded on the report).
• Outer-last generalizes to PERIMETER-LAST. An interior shared edge severs nothing — everything stays welded to the sheet skeleton until the band perimeter (ONE closed boundary) is cut last, severing the whole band at once. Whether the perimeter takes tabs is a post-stage decision, open with Jordan.
• Monotone band sweep — columns left-to-right / row bands bottom-to-top; after a perimeter severs a band the head never returns, so no rapid crosses a severed band.
• Sequences A and B (common lines/holes/outer; holes/common lines/outer) per the MetaCam reference block; C (part by part) IS the normal nest and refuses in those words.
• Shared edges cut on the nominal line with NO comp — no offset suits both parts at once; each part gives up half the kerf. That accuracy cost is recorded on every cut, never silent.

Eligibility is honest and strict (v1): the part's two extremes on the shared axis must each be ONE straight edge spanning the full extent, so translated edges coincide exactly. eligible() names the failing extreme in honest words; a circle, or nav209's lobed outline, refuses honestly. Every plan re-audits its own arithmetic (perimeter closure, sequence order, margins) with separate checks and raises on any miss.

The trade, in numbers (the show-the-trade ruling): per-sheet capacity common-line vs normal (gap zero packs more), total cut length (each shared edge cut once, not twice), and pierces stated honestly EQUAL — each shared edge pierces once, replacing exactly one part's outer pierce; 'connect holes to reduce piercing' is the ruled pierce lever, a later brick.

Server: GET/POST /api/jobs/<id>/commonline over the reviewed part (spacing from resolve_spacing, ran + ruled + source recorded); refusals are plain 400 in the engine's words, nothing persisted (the nestseq pattern); commonline.json persists; a re-plan in Part review or a geometry change DROPS it (the stale rule). UI: the ruled DEFAULT-DISCLOSURE block on the Nest tab — a chevron section, OPEN for a shop whose profile answered common-line yes, collapsed otherwise, never removed; axis / sequence / qty controls; the trade rendered in numbers; honest notes.

Proven: test_common_line.py 24 engine controls + app/verify_commonline.js 11 live checks over the public URL (scratch sketch fixture; nav209 untouched and honestly ineligible); all 14 suites green; shop.json restored unconfigured after the verify.

Honest limits: POSTING the shared-edge web is the next brick (emit knows only closed per-part boundaries today) — this brick plans and records; the canvas keeps drawing the normal nest until then; eligibility is the strict straight-full-span form (no partial-edge sharing, no connect-holes); perimeter tabs PARKED 2026-09-06 (Jordan: 'We need to ask others in the industry about this, being that we don't use common-line cutting.') — stays open pending industry input, not a shop ruling.