A couple of problems with our shader make it unfit for anything but the simplest cases. Because the brick pattern is computed with the modeling coordinates of the incoming object, the apparent size of the bricks depends on the size of the object in modeling coordinates. The brick pattern might look fine with some objects, but the bricks may turn out much too small or much too large on other objects. At the very least, we should probably have a uniform variable in the vertex shader to scale the modeling coordinates. The application could allow the end user to adjust the scale factor to make the brick pattern look good on the object being rendered.
Another potential issue is that we've chosen to base the brick pattern on the object's x and y coordinates in modeling space. This can result in some unrealistic-looking effects on objects that aren't as regular as the objects shown in Figure 6.6. By using only the x and y coordinates of the object, we end up modeling bricks that are infinitely deep. The brick pattern looks fine when viewed from the front of the object, but when you look at it from the side, you'll be able to see how the brick extends in depth. To get a truly three-dimensional brick shader, we'd need to add a third dimension to our procedural texture calculation and use the z component of the position in modeling coordinates to determine whether we were in brick or mortar in the z dimension as well (see if you can modify the shaders to do this).
If we look closely at our brick pattern, we also notice aliasing artifacts (jaggies) along the transition from brick color to mortar color. These artifacts are due to the step function causing an instantaneous change from 0 to 1.0 (or from 1.0 to 0) when we cross the transition point between brick color and mortar color. Our shader has no alternative but to pick one color or the other for each fragment, and, because we cannot sample at a high enough frequency to represent this instantaneous change at the brick/mortar border, aliasing artifacts occur. Instead of using the step function, we could have used the built-in smoothstep function. This function is like the step function, except that it defines two edges and a smooth interpolation between 0 and 1.0 between those two edges. This would have the effect of blurring the transition between the brick color and the mortar color, thus making the aliasing artifacts much less noticeable. A method for analytically antialiasing the procedural brick texture is described in Section 17.4.5.
Despite these shortcomings, our brick shaders are perfectly good examples of a working OpenGL shader. Together, our brick vertex and fragment shaders illustrate a number of the interesting features of the OpenGL Shading Language.