Does Code Decay?: Abstract
May 11, 1998
Statistical Engineering Division
DOES CODE DECAY?
Alan F. Karr
National Institute of Statistical Sciences
Research Triangle Park, NC 27709-4006
karr@niss.org
Developers of large software systems widely believe that these systems
_decay_ over time, becoming increasingly hard to change: changes take
longer, cost more and are more likely to induce faults.
This talk will describe a large, cross-disciplinary, multi-organization
study meant to study code decay as a scientific phenomenon, by
quantifying, measuring, visualizing and predicting it. The goals are to
identify its causes (both structural and organizational) and to devise
remedies. Emphasis will be on statistical issues associated
with code decay, using version management systems as the primary source
of data, on tools to describe and visualize changes and on assessing the
scientific evidence for decay.