ÖйúAƬ

While agreeing with every other point in Darrel Ince's article, I would question his suggestion that TeX is not based on software engineering nor validated. Formal methods of the sort that Ince has in mind are primarily conceived for coordinating teams working on projects.

TeX was written almost single-handedly by Donald E. Knuth using systematic techniques that Knuth calls "literate programming". The TeX command language is strictly closed; this is not a program that lurches wildly between versions. In concert with the program, Knuth wrote the definitive guide to TeX command language (The TeXbook) and a specific test suite (the TRIP test) that must be correctly processed by an implementation. The program has also been "validated" by thousands of users, who were offered an incrementing fee for any bug that they reported. The source code, command language and the ideas of literate programming have all been widely published.

Technical rider: the description above applies specifically to TeX the program. Knuth states himself that the program by itself provides features that are too low-level for everyday use, so TeX is the basis for various superstructure products to provide higher-level features. These are themselves written in TeX commands (macros).

The name "TeX" is, therefore, usually applied to Knuth's original user-version, formally called Plain TeX. Ince may have been thinking of the more widely-used LaTeX, which is subject to continuing development. I doubt, however, that the LaTex team would agree that they do not validate. Since the development process is completely open, perhaps Ince would broadcast (for example, on the info-tex discussion list) some recommendations.

R. Allan Reese

Head of applications

Computer Centre

Hull University