It just so happens that the knot gradients of a Bezier Spline can be expressed in terms of the adjacent control points:



Note: this approach is entirely synthetic!

Using the vector is reasonable, but what makes 3 a magic number?

Actually it is, but that's a topic for another time!