We are dealing now in questions of taste, for which tradition hath it that there is no accounting.
I do think, however, that one distinction needs to be made. The use of CIDTs has four important merits. They are derived algorithmically: no element of the arbitrary or historical enters into the determination of their values. Their implementation can be entirely automated. An implementation using lub-seeking binary search at execution time is highly efficient. The scheme itself is perspicuous, easy to take in and understand. These seem to me to be formidable merits. It is, of course, possible to choose a command-keyword set or the like badly. How not? Such bad choices are indeed common. Pointing to their deficiencies is not, however, an appropriate criticism of CIDTs. A birthday cake may have an emetic Paris-Green frosting; but the froster, not birthday cakes generically, should be criticized when this is the case. (The rhetorical device of confusing bad exempla with a scheme itself in fact has an impressive Greek name.) John Gilmore, Ashland, MA 01721 - USA
