Experiments aimed at gaining insight into the complexity of object-oriented code maintenance are described in this paper. The authors quantitatively summarize their experimental results, comparing and contrasting them to those of others. As the authors acknowledge, the collective evidence suggests that the questions for which an answer was sought are probably incorrectly formulated: inheritance depth may not be directly correlated to maintenance cost.
This work attempts to simplify the management of a daunting experimental space, accepting and acknowledging the limitations imposed on its results. The product is an experimental report that discusses an interesting methodology, and the results and conclusions are of preliminary value.
Considerations of the experimental space aside, quantifying code maintenance cost must always deal with at least another difficult problem: quantifying source code complexity. Perhaps the route to quantifying code complexity is indirect, progressing through the assessment of maintenance cost first. As difficult as experiments are in this context, they can provide evidence of the usefulness of software metrics. This work could be construed as an early, and perhaps not fully intentional, contribution in this direction.