Opening the viewer found in ten seconds what the contact sheet had hidden
through two rounds of refinement: the bellows was cantilevered off a single
post driven through its middle, with four fifths of a door-sized leather bag
hanging in open air. The 300px thumbnail is too small to see what is under a
thing. A viewport you can orbit is not.
The support check passed it, and was right to by its own lights. The post
reaches the floor and the nose touches the hearth wall, so the assembly IS
connected to the ground. Connected is not the same as looking like it could
hold itself up, and looking like it can is the entire rule. The check
answers a question about topology; the rule is about weight.
So the bellows gets the bench it always had in life: two bearers let into the
forge wall at one end -- let INTO, a third of the way through, because a
bearer that stops at the wall face is a bearer resting on nothing -- carried
on legs at the other, with a rail across their tail. Six plain boxes, no
rotations at all, which after three two-rotation mistakes in this one model
was the point. The post moves out beyond the tail where it stops spearing
the bag.
Two more things came out of the same look.
The lever ran the wrong way -- from the post out into empty air, away from
everything. The end a smith pulls is the end he can reach standing at the
fire, so it now falls back over the bellows towards the hearth. And it had
nothing to pull: a rocker with no connection to the top board is a stick
resting on a post. There is now an iron link rod through the board, which is
also the only thing tying the two halves of the assembly together to the eye.
And the leather. The last round widened it so it would stop hiding between
the boards, but width and length were the same parameter, so it grew a third
longer in the same breath and the bag ended up running well past the boards
it is nailed to. Leather cannot outrun its boards -- they are what it is
fastened to. The two scales are now separate and the hide is wider (1.18 to
1.33) and shorter (0.82 to 0.9), which is the real thing and also spreads
the five end caps further apart than the old scheme managed, since they are
now staggered on purpose rather than as a by-product of the width.
+72 triangles, six boxes. 34 models, 29 094 triangles, all checks pass,
0 floating cases, GLB round-trip clean, Conformant, build clean.
CRT
fb092b8ef0