End-to-End Arguments in System Design
Saltzer, Reed, and Clark formulate the end-to-end argument: many functions can only be implemented correctly at the endpoints, with lower layers serving mainly as performance optimizations.
Saltzer, Reed, and Clark formulate the end-to-end argument: many functions can only be implemented correctly at the endpoints, with lower layers serving mainly as performance optimizations.
Using the KWIC example, Parnas shows that good modular decomposition hides change-prone design decisions rather than mirroring processing steps.
Brooks distinguishes essential from accidental difficulties in software engineering and explains why no single technology can deliver an order-of-magnitude productivity jump.