<GitHub>
[solvespace] jwesthues commented on issue #246: `RANK_MAG_TOLERANCE` is absolute. It should obviously be relative to the bounding box of the sketch; but we have enough dynamic range that it never bothered anyone enough to fix it. https://github.com/solvespace/solvespace/issues/246#issuecomment-298145162
<GitHub>
[solvespace] jwesthues commented on issue #225: It's the number of digits of precision that we lose while solving; but we need to define the condition number as the ratio of the maximum singular value to the minimum **non-zero** singular value. I suspect that your condition number is artificially big because your sketch has unconstrained DOF, and each unconstrained DOF introduces a singular value that's ideally zero.... https://github.com/solve
<GitHub>
[solvespace] jwesthues commented on issue #225: Okay, got it. So you've got 32\*4\*2 rectangle vertex coordinates, minus 31\*2 unknowns for the point-coincident at the corner, minus 32\*2 for the lengths, minus 32\*2 for the parallels, minus 32 for the perpendiculars. That's our 34 DOF.... https://github.com/solvespace/solvespace/issues/225#issuecomment-298188156