Most programmers these days write in object-oriented programming languages, such as Java or Python. This category also includes much Perl 5, C++, and Objective-C code. It even includes much Javascript, which work on the Document Object Model. Such code is organized as a collection of (classes of) objects. You get a new object by demanding a new, blank element of some class, or by demanding that some other object be cloned. So why do I write functional programs, not object-oriented programs? Because these are easy and hard to extend in different ways.

With object-oriented programs, it’s easy to extend them by adding a new class of object. And it’s easy to add a function to one class… though you then have to consider all its children. It’s hard to add a new feature in a broad way, like serialization or access control: you have to change each of your class definitions, likely every file in your program. So nouns are easy, but new verbs are hard.

With functional programs, it’s hard to add a new sort of noun to the world, but easy to add a new sort of verb. Since I tend to program in Unix-like environments, this is nice: I know generally what sorts of noun I’ll be dealing with. I write programs to add new verbs to my vocabulary.