Except when you large monolithic codebase is composed of a whole bunch of screwy distributed components that have been thrown together haphazardly. And, yes, I

Except when you large monolithic codebase is composed of a whole bunch of screwy distributed components that have been thrown together haphazardly. And, yes, I know you can't have a "monolithic codebase" with "distributed components". It's just that it's actually all the same dish with a different sauce. Distribution of processing doesn't create better code bases -- it's decoupling of the computation that does it. You can decouple code and a very large number of ways. It's also just as easy to write a rat's nest of confusing spaghetti with components that hold each other's state as it is to do it without "components" (I speak from experience on this front, as much as I hate to admit it).