[kandelabr]

I'm not an expert on meshing or anything but I've been thinking about snappy and I wonder if any of that is of any value.

First, how many cells/points are moved when snapping to surface? For example, if you have a stack of cubes and move one vertex of the topmost down by a significant amount you get a bad cell. But if you spread the movement among many cubes in the stack you limit the deformation of each cube. Does snappy do that and how deep does snapping deform cells?

Then, about the almost useless layer addition: what if instead of first snapping and then adding layers one would first offset, snap to offset surface and then extrude layers?

So to elaborate a bit more:

1. Offset STL surface by overall thickness of all layers. If that would produce self-intersecting geometry one could either limit offset amount or create illegal geometry and then clean up after offsetting (?).
2. Act as if the offset geometry is the original one and make all the snapping stuff on it.
3. Extrude a single layer from snapped mesh to original surface. That would ensure all boundary cells are prismatic.
4. Divide boundary layers to user's specifications.

What do you think of that, is it possible or any better or am I missing something?