> But it seems like a falsehood to suggest that it's not possible to understand one level of abstraction without knowing the full stack; isn't that why we have

> But it seems like a falsehood to suggest that it's not possible to understand one level of abstraction without knowing the full stack; isn't that why we have abstractions in the first place? Any non-trivial abstraction will have some leaks. You can memorize the problem cases and usefully use the abstraction without understanding the source of the leaks. Partial understanding can be sufficient. Fully understanding an abstraction probably requires you to understand all the levels of the stack. Is it actually useful to know the stack top-to-bottom like that? Not always. As you note, that's why we cover up the details in the first place, so that we don't always have to keep the underlying implementation in mind.