If kingdoms were really lost for want of a single horseshoe or, as Pascal conjectured, the entire face of the earth would have been changed had Cleopatra's nose been shorter science and engineering as we practice them would be impossible.
Without what Bertrand Russell called "approximately separable causal chains" some kinds of analysis are precluded, and we have reached that stage for the pipelined execution of a sequence of z/Architecture machine instructions. The speed of a single instruction in such a sequence cannot be analyzed independently, and we delude ourselves if we suppose that SR Ri,Ri can be shown to be globally faster than XR Ri,Ri (or vice versa). All we can do is show that one or the other of these results obtains in particular, fully specified circumstances; and we cannot often export such a result, use it in different circumstances, with any confidence. We can live with this, give up single-instruction timings; but it will apparently be very difficult for some of us to discard old epistemes. John Gilmore, Ashland, MA 01721 - USA
