PCA applies an orthogonal linear transformation, while FA uses a series of coefficients to scale a sequence of functions, which are then integrated. They are similar in use but very different in method. Calling one a generalization of the other seems misguided?