Originally it was for more convenient usage of homogeneous coordinates
in shaders themselves, but it is actually not needed:
* When passing three-component vector (3D position) to `vec4`, the last
coordinate is implicitly set to `1`, thus there is no need to pass it
explicitly in each attribute.
* On the other hand, when passing three-component vector (2D position
in homogeneous coordinates) to `vec3` with Z explicitly set to `1`,
it still needs some swizzle magic to extend it to `vec4` gl_Position.
Passing it as two-component vector results in nearly the same magic
while saving precious memory.
* In shader uniforms (projectionMatrix makes more sense than projection
alone)
* For underlying types for SceneGraph transformation. It is already
used in Drawable::clean() as transformationMatrix, so why not use it
also in AbstractFeature::clean(). Moreover, clean() could be in
future also done using something else, this helps to distinguish the
type just from parameter name.
* In Physics shapes - applyTransformationMatrix() (as it could be in
future also done using something else).
Vladimír Vondruš