Beware of survivor bias. Most old projects failed, like most new projects. But why is it that the rate of failure in programming projects hasn't fallen? I've

Beware of survivor bias. Most old projects failed, like most new projects. But why is it that the rate of failure in programming projects hasn't fallen? I've heard it said that the computing world in general operates on the edge of acceptability. As soon as the browser became fast enough to become an applications platform, we implemented the browser in the browser; as soon as good command-line interfaces develop, we replace them with bad windowing ones; as soon as dumb phones become a mature technology, we all move to highly experimental smartphones. This is probably most severe in programming itself, where our field's most free to make its own choices. The percentage of projects that fail remains more or less constant as our level of experience grows -- because, as we become more experienced, we move to more complicated models. Any modern programming shop could churn out successful COBOL waterfall projects at a rate of two or three a year; so, as we got good at waterfall and old languages, we all moved to Agile and OO. Then, once we developed dangerous amounts of expertise in Agile and OOP, we moved to the Web (and lost most of our ability to use IDEs). And now, with even Web programming in danger of becoming a mature discipline, I'm hearing a lot of talk about Haskell. Why does this happen? I don't know -- because I don't feel the temptation to do it. I program in C++, and I carry a feature phone...