Tool Implementation

Abstract

‘RESEARCH tools’ is a name given to a peculiar kind of software written to aid in experimental computer science. An analogy is implied with more usual scientific equipment likemicroscopes or spectrometers, tools used not to construct, but to probe and study. There is also an implied disclaimer: research tools are often throw-aways. Only the experimenter can use them, and when the experiment is finished, they are very difficult to use again. Thus research-tool software is fragile, hard to use, poorly documented, and full of failures waiting to happen. Part of the reason is that the tools, like Topsy, “just growed.” The experimenter starts out needing something simple and writes a quick-and-dirty programto do it. As the work continues, the program is augmented and modified, its deficiencies papered over, until it is a rat’s nest of bad code. There is never time to document or to rewrite. Failures appear each time the tools are used in a new situation (or even in attempting to repeat an old experiment because they’ve changed since it was first run), leading to further ‘development’ that makes things worse.