When you have defer, you still need to write code you'll be deferring. That's what the destructor is, no more. `defer` is strictly additional overhead.

Furthermore:

1. If you're just composing in destructible objects the compiler does the work for you, if I create a struct which contains a File, i don't have to do anything, rustc will make my struct implicitly Drop and release the `File`. If you do that in Go... you have to remember to write a deferrable method which'll close the underlying file.

2. You're only defining a type once, and you're thinking about its invariants right there, it's a lot easier to remember those invariants when creating the type than every time you're using an instance.

Unless you have linear structural subtyping and the compiler forces you to do it. Then it's fine.