Following Clarification about quad node ordering, I made some experiments and here is a snapshot
This image shows two surfaces where the colors correspond to the Y component of the normal computed by the vtkPolyDataNormals filter with ComputeCellNormalsOn and ConsistencyOn.
As you can see, the normal (i.e. orientation) of the polygons (including quads) of the surfaces have been computed differently (red if Y positive normal, and blue is Y negative normal where Y axis points towards me/camera). From the same point of view, polygon node ordering is on one hand clockwise and on the other hand counter clockwise.
Per latest point and per vtkQuad documentation, it seems that the polygon orientation is constrained by VTK but I don’t understand why and how.
Indeed, it seems that I can have a surface with polygons having different orientation. It seems that a VTK filter can order nodes of disconnected polygons in a different clock orientation. It seems that whatever polygon orientation I have, VTK can still render the surface without any warning.
Here I don’t pretend at all that it is recommended to have heterogeneous orientations in some surfaces. This is not at all the topic. The topic is really to understand if VTK has a constraint on polygon node ordering and if so, how this constraint is enforced especially knowing that a VTK algorithm can compute different node ordering.