Experimental Evaluation of Software and Systems in Computer Science

Punched cards were already obsolete when I began my studies at the Technical University of Munich in 1971. Instead, we had the luxury of an interactive, line-oriented editor for typing our programs. Doug Engelbart had already invented the mouse, but the device was not yet available. With line editors, users had to identify lines by numbers and type in awkward substitution commands just to add missing semicolons. Though cumbersome by today's standards, it was obvious that line-oriented editors were far better than punched cards. Not long after, screen oriented editors such as Vi and Emacs appeared. Again, these editors were obvious improvements and everybody quickly made the switch. No detailed usability studies were needed. "Try it and you'll like it" was enough. (Brian Reid at CMU likened screen editors to handing out free cocaine in the schoolyard.) Switching from Assembler to Fortran, Algol, or Pascal also was a no-brainer. But in the late '70s, the acceptance of new technologies for building software seemed to slow down, even though more people were building software tools. Debates raged over whether Pascal was superior to C, without a clear winner. Object-oriented programming, invented back in the '60s with Simula, took decades to be widely adopted. Functional programming is languishing to this day. The debate about whether agile methods are better than plan-driven methods has not led to a consensus. Literally hundreds of software development technologies and programming languages have been invented, written about, and demoed over the years, only to be forgotten. What went wrong?