Also shows why flexibility of the micro-kernel approach is good.
Similar goals and approach (at face value) as first generation. What is the key difference in better results?
