[jacksmith21006]

There is a number of things think Fuchsia or Zircon more specifically have correct.

Couple of big ones is first using a micro kernel. We really need this at this point for granularity in a MP setting so cache will be hot.

So a driver can be pinned to a processor so working with cache instead of going to memory.

Then with the ability for a thread to sleep based on a montitor/MWAIT on a specific processor.

This way you can basically create a pipeline that is super fast and uses memory to communicate.

The other really smart thing was how drivers are using a Vdso and therefore mapped into all process address space.

Then the obvious is using a memory capability approach which removes a lot of marshaling that usually has to happen by the kernel.

There is going to be a lot of interesting ways to leverage Zircon.