"DBus Embedded" - a clean break

[Havoc Pennington]

btw, I'd highly recommend that any benchmarking include the following
cases, just to show the different tradeoffs:

1. blocking calls

 for 1 to N {
    send message
    wait for message to come back
  }

2. batched calls

 for 1 to N {
    send message
 }

 for 1 to N {
   wait for message to come back
 }

3. blocking and batched calls with tiniest possible messages and with
pretty large messages

Large messages reveal libraries that make too many copies or
allocations, which is invisible with tiny messages. The batched call
benchmark profiles how apps should work, and the blocking benchmark
profiles how they often do work.

Havoc