Sunday April 16, 2006

The dangerous myth of "computer science"

"Computer science" is not a science, not in the purest sense, which is a system of inquiry for the determination of general laws. Instead, as Paul Graham has written, "Computer science is a grab bag of tenuously related areas thrown together by an accident of history." Now, myths can be benign, even comforting, like the myth that Texans are cowboys, or that New Yorkers are sophisticates. But myths can also be dangerous.

"Computer science" is a dangerous myth when it leads us to mistake effort for progress. We tend to think that a piece of software that was written five years ago, in an "old fashioned" language, is somehow less valid than one written today, in a newer language, because the one written today is somehow the beneficiary of a more advanced "science." This is absurd, because down at the level where computer technology has some claim to be a science, both programs will draw from the same set of machine instructions.

I once worked for an outfit that wrote the same piece of enterprise software three times, each time in a more "modern" fashion, using different languages, and each time the gain in functionality was minimal. If they had spent those hundreds of man-years on making improvements that customers actually cared about, instead of following the latest fashions in "computer science," they would have been wildly successful. Instead, they alienated what had been a very loyal customer base, and never made much money.

I think we would all be better off if we began to regard our profession as more of an art than a science. Artists don't throw away the past -- they build on it.