In current specification for transform, different operations possible are translate, scale, rotate, skew and matrix (all being some form of matrix). The elements in the matrix are constant numbers or some constant function. By allowing these elements to be functions taking x, y (and z in case of 3D) coordinates of current point being transformed as arguments, could result in producing more complex transformation (3D effects in a 2D structure, more complex transformations like projections in 3D structures). Essentially, making the transformation dependent on coordinates of each point on the structure being transformed.