New metrics needed for new generation: lines of code, function points won't do at the dawn of the graphical, object era - computer-aided software engineering: includes related article on Windows test tools and a glossary on metrics for object oriented design - Buyers Guide

Software Magazine, May, 1992 by Jessica Keyes

Since most client/server systems of the future and all graphical user interface systems are object-based, the IS world is beginning to take notice of this type of research.

Kemerer proposes a series of six metrics to measure the depth and breadth of object-oriented design. (See box, this page.) He is collecting statistics to verify his proposal.

Object-oriented metrics are specifically oriented to object-oriented methodologies, which are quite different from conventional methodologies. Kemerer's metrics measure depth of inheritance, which differs from conventional measurements.

"The notion is to try to go after those things that are different about the object-oriented approach. The one that is the easiest to explain to most people is the notion of inheritance," he said, where one class of objects inherits properties from a higher class. "Our metric is to measure depth of inheritance. In this way, we can determine to what degree people are using inheritance."

The goal is to address the optimal mix between complexity and usability. From a human perspective, 10 or more of steps, variables or anything, becomes too complex.

When a programmer does not use inheritance, he or she is not taking advantage of reusability, therefore negating the productivity gains of the object-oriented technique. When the programmer "goes really deep," according to Kemerer, "this may also be bad since it will be hard to test it. Indeed, it may get to be too much for one person to keep in mind."

Even though Kemerer is a proponent of this new form of measurement he, like Rubin, is still keen on function points. "Measurements could co-exist," said Kemerer. "Measurements might address two parts of the same problem. One is size; the other is design complexity."

There are many valid measurements available, but research has shown that productivity measurement is used by a low percentage of IS organizations.

"The efforts that have been put into measurement are from consultant firms that want to sell measurement tools, or they're done by organizations that want to prove that a new approach will really pay off," commented Chris Gane, president of the New York City-based Rapid Systems Development and co-developer of the Gane/Sarson developmental methodology. Gane is also a consultant to Bachman Information Systems located in Burlington, Mass.

Rapid Systems Development markets HyperAnalyst, a software tool that automatically details what a methodology should look like depending on a given set of complex circumstances.

No doubt many IS organizations are put off by the expense of hiring a consultant, no matter how much potential there is in their methodologies for TQM. Perhaps many have no time or are simply not interested.

However, Lou Mazzucchelli, chief technical officer for Cadre Technologies, Providence, R.I., predicts that enforced measurement is on the horizon. "It could come from places you wouldn't expect -- like regulatory lawyers," he explained. "Lawyers could force the system to evolve such that you are required to measure, to prove you have done things in an appropriate way."