I would not respond to this nonsense except that Al has chosen to broadcast it to so many newsgroups. The author is woefully ignorant. To pick a single point, he says in his first sentence: "Response surface methods in DOE entail fitting a nonlinear model to an observed response, usually some form of a polynomial function."
(1) Polynomials are linear models -- the fact that some of the terms are power terms has nothing to do with it. He later on proposes an analysis method (SVD) which depends on this linearity, and of course he is confused about what SVD does.
(2) One can treat non-linear models, such as y=a*e^(b/x), but only one point at a time, since the optimal design depends on the coefficients. This dependency is absent when the model is linear.
etc.
Bob is the President of eChip, and one of the old hands in the Design of Experiments field.
Ron Schoenberg is, of course, not ignorant, as anyone familiar with either his papers or his accomplishments will attest. At my request, he wrote an article designed to stir up a little controversy, and, as you can see, succeeded perhaps even a little better than he intended.
Bob himself allowed in a subsequent note that Greg Piepel's letter was a more accurate reflection of his own feelings than this note. We include this to illustrate that Design of Experiments is not a dry academic field. Bob feels passionately about these topics, and he is not the only one. For example, we got a few comments from people who said it was about time that someone wrote an article about situations where screening was not the primary objective.