ITER is a global scientific collaboration to prove the feasibility of fusion energy from magnetically confined plasmas on an industrial scale. Construction of the ITER facilities is underway at Cadarache in the south of France. To finish detailed ITER design, operate and interpret ITER, and thence to design future fusion reactors, it would be desirable to have software to model the ionised plasma and its interactions which has a thirty-year-plus life span, capable not only of use but also of significant modification until c.2050. The modification requirement arises because the baseline for designers is currently a set of zero-dimensional codes, ie. where there is no explicit modelling of any plasma spatial dependence, to be augmented with more sophisticated plasma models as time and consequently computing power, and ITER operational experience increases. Most software users will be designers who want the ability robustly and quickly to explore physical and engineered parameters with minimal knowledge of the inner workings of the software. Robustness implies the use of tight-coupled numerical solver algorithms in the context of changing machine architectures. Speed of software execution will be important both for feedback to the designer and when ensemble calculations are required. Software engineering techniques for achieving the above goal will be discussed. The most promising appear to be the use of layered software and hardware, object-oriented techniques, good programming style[1] and code written to read like a book.

[1] Arter, W. and Brealey, N. and Eastwood, J.W. and Morgan, J.G "Fortran 95 Programming Style", Report CCFE-R(15)34, http://www.ccfe.ac.uk/assets/Documents/CCFE-R\%20(15)34.pdf.