The CAD systems in use in the knitwear industry are developed by manufacturers of industrial knitting machines for use with their own machines. (In addition to the machine builders, Minima, a software house based in Reutlingen, produces a CAD system for programming knitting machines, which is compatible with the machines produced by all the flat bed machine manufacturers: Universal, Stoll, Shima Seiki, etc.) These CAD systems are designed to increase the ease and efficiency of programming, and not at all to support the invention of the designs themselves. They are primarily visual programming environments, and are intended for the technicians to translate completed designs into runnable programs.

Why have the producers of CAD systems for knitwear neglected the designers? Both the objective characteristics of the industry and the available technology, and the attitudes of the system developers play a part, but we argue that the former are much more significant. Nevertheless the technology has progressed to the point where supporting the designers has become a relevant issue, and the development of systems that suit knitwear designers is held back by the attitudes and beliefs about knitwear designers and CAD systems held by system developers, knitwear company managers, design teachers and the designers themselves.

Knitting machines need to be programmed somehow, so the machine manufacturers must provide some way to do it, and the quality of the support provided for the programming is a major selling point for the machines. This is the driving force behind the development of better CAD systems for knitwear. The knitting machine companies have constant contact with technicians using their machines, and can see how well the technicians can learn to use their systems in the training courses they provide. Conversely, having technicians who can program knitting machines well and efficiently is a clear economic necessity for knitwear companies, who will usually pay for them to attend training courses for new systems. This aspect of the business has not changed since the days of metal punch cards, when the division of responsibility between designers and technicians was absolute.

The state of CAD technology also influences the emphasis given to supporting technicians rather than designers. Computer control of complex machines is a well worked out area of engineering competence, and producing good programming environments is very much easier than producing good CAD systems for informal design work. The first reason is that the discrete nature of knitting patterns (which are matrices of smallish numbers of stitches) makes it relatively easy to represent formally stitch structures and algorithms for producing them in a variety of different ways. Universal and Shima Seiki use colour coding; Stoll uses symbols to represent individual stitches as part of a language derived from BASIC, and has introduced an iconic representation of predefined stitch structures. The second reason is that knitting machine programming is a job that fits comfortably into the task-order constraints imposed by relatively simple software systems, whereas design doesn't. A third is that unsolved hard problems are involved in providing good CAD support for sketching and geometric design that permits imprecision and progressive refinement of details, and avoids premature commitment and hidden consequences of decisions (Tovey, 1992, Scrivener, 1993). The analogy to other design tasks where similar patterns of computer use obtain supports the view that the relative difficulty of supporting designers is largely sufficient to cause the existing imbalance in computer use (Tovey, 1989, 1992).

One almost entirely male group of computer programmers and mechanical engineers develops systems for another almost entirely male group of technicians, but we think the sex division is less significant than a personality type division. The system developers understand the thought processes and priorities of the technicians because both are programmers, but have difficulty understanding the attitudes, values and working methods of the designers. One illustration of this is the colour coding of stitch types used in the Universal and Shima Seiki CAD systems: they are effective formalisms for visual programming, but (as the first author has observed herself) they are unpleasant for people trained to be exceptionally sensitive to colour combinations. Computer scientists need detailed study to understand designers, as we know from our own experience (Eckert and Murray, 1993), as well as that of others (Scrivener,1993) and can go seriously wrong. Ashby (1992) observes that some CAD systems for textile design try to force designers into using alien working methods; while Waddell (1992) comments that some fashion design CAD systems are based on a misapprehension of the nature of the design process, often coming from a failure to understand what the designers' sketches are actually for, with the result that fashion designers find them useless and feel themselves excluded from the technological age.