What remains from this discussion is that how should be handled the edge of a primitive depends on various considerations; some being specific to some use case and/or hardly known from the structure itself, some other being not clear if it's actually better a way or another (which would induce some deep research to determine).
So I think the best would be to allow configuring MeshlinePolicy::Policy (and other settings) on a per-something basis:
And then configuring some best-effort profiles per use case (for instance PCB, RFIC, Waveguide, etc.)
Also the paper Optimum mesh grading for finite-difference method suggests the optimal smoothness factor is around [1.2 ; 1.4].
What remains from this discussion is that how should be handled the edge of a primitive depends on various considerations; some being specific to some use case and/or hardly known from the structure itself, some other being not clear if it's actually better a way or another (which would induce some deep research to determine).
So I think the best would be to allow configuring
MeshlinePolicy::Policy(and other settings) on a per-something basis:And then configuring some best-effort profiles per use case (for instance PCB, RFIC, Waveguide, etc.)
Also the paper Optimum mesh grading for finite-difference method suggests the optimal smoothness factor is around
[1.2 ; 1.4].