Planned - short term:
- taktuk pilot lib ?
- wait command ?
- added test for connector propagation (fall back on flat on failure)
- print_tree does not work correctly when executed on non root node

Planned - long term:
- good algorithm for adaptive window

Given up (see construct tests for some of these cases):
- efficient transfer of taktuk engine - actually not efficient (using the way I
  had in mind)
- sendv/recvv of different sizes - given up, not meaningful to TakTuk
- reduction of the number of concatenations - no easy gain still under
  investigation
