I have come to believe that static typing zealots haven't worked seriously on a large variety of (shipped) software projects. I certainly was one when I was in

I have come to believe that static typing zealots haven't worked seriously on a large variety of (shipped) software projects. I certainly was one when I was in school. Static typing is a good choice when the system you are implementing is already defined. It's great for implementing a well-defined algorithm or protocol, or a well-studied domain like game engines or financial exchanges or rocket ships. Basically, wherever correctness is necessary and possible. Static typing is a bad choice when the system you are implementing is largely undefined, or actively evolving. That is, "the rest" of the software, where "software correctness" is undefined. Which includes things like a business, a new video game, a website, and scientific research. Because you will waste time building an ontology of types that can't possibly be known at the time, or worse, constrain the natural evolution of the project. CS students often get a warped perspective of software, where everything is well-defined by their professor or some textbook or RFC or some other smart bloke, and they just have to implement it. In reality, the average software project is not this way. The end result is Enterprise Java (formerly C macro hell). You get a bunch of CS grads who think static typing is the only answer working on some fuzzy business logic with Java. The conclusion is an unspeakable monstrosity of bad abstractions (and job security).