Thanks for elaborating. To counterpoint a little... from a developer's perspective, while the protocol stack model is a very useful conceptual tool, it's not necessarily the best recipe for actual coding. Message passing across abstract boundaries imposes costs. The point at which it becomes "too much" may be a judgement call about engineering tradeoffs without a "best" answer. There may be analogues in regulatory policy. Of course, I'm not denying that "too little" structure and abstraction generally results in buggy, unmaintainable spaghetti.