An Empirical Comparison of on-Line and off-Line Debugging

摘要

The experiment compared the program debugging (checkout) performance of programmers using a Time-Sharing System (TSS) with the debugging performance of programmers using a simulated closed shop. Twelve programmers participated in the study. Each programmer was given two problem statements and was asked to write a program to solve each. One proglem required a program to interpret and solve algebraic equations. The other problem required a program to find the single path through a 20 x 20 cell maze represented in the computer by a 400-entry table. Six solutions (programs) to each problem were debugged on line using TSS and six were debugged off line using a simulated closed-shop system with a desk-to-desk turnaround time of two hours. Performance was measured in terms of man hours to debug and central processor time used in debugging. Programmers who debugged their algebraic interpretation programs on line used slightly fewer man hours and about three times as much central processor time as did programmers who debugged these programs off line. Programmers who debugged their maze programs on line used about one-third as many man hours and slightly more central processor time than those who debugged their maze programs off line.