kerfmaster pin list · The spec — what we build to
recordStates what the code does today. Auditable against the code — this is the only type that can go stale.
Content last changed 2026-08-07 — computed from the item itself, not typed.
Contract — design stage
This contract describes the code as it runs today.
takes —
makes —
fails if A second jig built to reach across the standoff; a throwaway hole in a plate face that is not covered by another plate.
Contract last changed 2026-08-07 — computed, not typed. Dated separately from the text above, so neither date can speak for the other.
mvt_jig.py builds the installation template for option C only (Aristide, 2026-08-07). It writes viewer/mvt/jig_green.dxf, plates_ring_tri.dxf and jig.png.
I argued that option C's plates sit 0.5 in apart on spacers, so a flat green could hold only the layer it lies in and a second jig would have to reach across the standoff. That was wrong, and the answer was already in the spacer siting the customer preview publishes: in option C both the ring and the triangle stand on spacers fixed back to the wall, running clear past the cloud on the way. Every fixing in option C is therefore a hole in the wall, all six of them in one plane, and one flat template locates the lot. wall_sites() re-runs build_mvt_preview.assess('C') live rather than copying the coordinates, so the template cannot disagree with the sheet the customer approved.
The cloud nests in the window it was cut from, so the window matches the plate because it is the plate's own cut line. It is held by straps: a tie passes through a hole in the green, across the front face of the cloud, through a second green hole, and back along behind. Both holes are in sacrificial material and the finished plate takes no hole at all.
This replaced the paired-hole scheme (one hole in the green, one in the black) for a measured reason. Aristide's rule is that a hole in the black is fine only where it is never seen, and the cloud's never-seen area is just the 19.47 in² the ring and triangle cover — confined to a 10.8 × 7.1 in patch in one corner of a 28.7 in plate. Four tie points inside that corner cannot hold the plate flat in its window. A strap has no such constraint, because both of its holes are throwaway.
A candidate strap must lie wholly on metal and bite it: the middle of the chord has to be at least MIN_BITE 0.30 in inside the boundary. Without that test the search returns chords that clip a corner — covered by metal the whole way, gripping nothing, and the tie cants on the lip. Straps are scored on the distance between the two green holes (that is the loop the tie has to close) and spread by the separation of their midpoints. Measured on this artwork: 6 straps, spread ≥ 6.70 in, spans 2.17–6.21 in, so a 13.4 in tie reaches the worst of them.
The ring and the triangle are not in the template. Nothing covers either of them, so neither can take a throwaway hole — and neither needs one: they are located by the wall holes this template marks, and they go up after it comes off. They are cut as ordinary parts; nesting is a later stage.
Six 0.375 in marker holes sit at the spacer centres. All six land on green, none inside the cloud window. The tightest is one ring spacer 0.353 in from the window edge, leaving 0.165 in of green round it — reported every run rather than quietly moved, because moving it would move the spacer.
Hole size is Jordan's two brackets: ≥ ~0.307 so the pierce gets 0.150 in of clear air inside the hole, ≤ 0.75 so the slug drops through the slats and needs no retention tab. 0.375 sits inside both, so there is no pierce exception list and no slug tab.
Sheet 33.000 × 25.289 in, inches throughout, well inside the 80 × 160 bed. Worst fit deviation 0.00141 in, worst fit-induced kink 0.029°.