When you use a Triangulator on some surface that has a transformation matrix, the matrix is always applied. I propose that FME respects the matrix and leaves it intact, performing the triangulation on the local coordinates of the geometry. Or at least make it optional.
Now, I realize that this wouldn't work for aggregates/multis, because these might have a "parent" matrix and "child" matrices, whereas the result of the Triangulator would just create one big mesh. That could be dealt with in one of the following ways:
- For these types of geometry, just apply the matrix (keep current behavior)
- Only apply the child matrices, but keep the parent matrix intact and transfer it to the output mesh
- Triangulate all non-aggregate/multi child parts and leave all matrices intact, which means that the output will not be a mesh but remains an aggregate/multi
I'd prefer option 2 or 3, but of course, this could also be another option which can be set on the Triangulator.