I can't recall if someone authoritative like stewer ever gave info on internal triangulation or not.

I have to get to work, but I have an experiment in mind that can perhaps give some info on that.

Suppose internal triangulation does happen. Given a quad, there are two diagonals available to minimally triangulate. Another possibility would be that the quad is made into four triangles meeting at a new hypothetical centroid. (I doubt this but we can test for it.)

Now if you were going to build the triangulation code, you'd have a choice of choosing the diagonal randomly, or choosing it based on vertex order. The latter is more likely. A third, very unlikely choice, would be to use world coordinates to make the choice.

All of these have differing implications that can be tested.

Suppose instead of using the Poser primitives, I build some procedurally, with great precision.

Suppose I build, from identical vertices in world space (A, B, C, D), four different quads:

ABCD

BCDA

CDAB

DABC

If I were to simulate and render these, and I get different outcomes, even though the four points are the same, then we know vertex order is a factor, and further, we'll probably get some evidence of triangulation choice method for one diagonal or another.

Further, we can rotate these so that they align with the other variations. This will reveal whether or not it is model space or world space that influences things. Doing this requires great precision (rotating about the center of the square very carefully) so I would not attempt this with dials. I would do it procedurally, via a script.

We'll see what that reveals.

 

---

Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)