Why MVP matrix in Tessellation Evaluation Shader?

I have seen some examples of how the hardware tessellation can be performed, and in all of them the MVP matrix is applied in the tessellation evaluation shader.

Wouldn’t it be better to just transform the patch control points early, say in the vertex shader?

If the “tessellated vertex position” can be expressed as a function of the control points and the u,v,w parametrization (which would be the common case?) I can’t see why one would wait with the MVP transformation until the TES.

OK, I could just try it myself, but I’m curious about your thoughts.

(PS. If anyone knows why the tessellation parameter space (triangles, quads, isolines) is declared in the TES and not in the TCS I would like to know that too. )

You don’t have do wait until the tesselation stage; but it is often convient to have the patch vertices in a local coordinate system in the tesselation stage.