The second method of modeling an object is by polygons. Generally, 3 or 4 vertex polygons are used. Triangles are more often used because all vertices of a triangle are coplanar, so the polygon will have only one normal, which greatly simplifies some rendering computations such as shading. The scene designer will construct an object one polygon at a time, and then define the surface texture and the color for each of the polygons.
The motivation for defining an object by polygons is flexibility of the object representation. Each polygon is represented by a fixed set of points and edges joining the points. These points can be arbitrary positions in 3D space, so deforming an object is a relative simple task (e.g. a filled milk carton's side can be slightly warped).