- what to do with IV, should it be a cipher thing or a protocol thing?
  ( currently swaying toward a 'cipher thing' since not all ciphers
    maintain IV information and this way the protocol engine does not
    have to worry about it; also another bonus is that the IV does not 
    have to be passed to the cipher engine on every packet )
  [ BTW, current implemntation is 'cipher thing' ]

- should the callback be issues on every job execution?
  a) if the job fails, or is finished, or is scheduled in async mode the
       callback function is called.  The advantage: if there is any code
       to be executed after a job is finished it must only be placed 
       in the callback and not also after the 'execute_job'.
  b) if the job is scheduled in async mode the callback function is used,
       otherwise (failure/software processing) the result code must be 
       checked and no callback is called.  The advantage: if the engine
       is a software one, there is less stack waste for nested/sequential
       engines.
  [ current implementation uses 'a' which optimizes for less code duplication
    and better performance for async engines ]

- write more tests:
  - MD5 known values test
  - 3DES known values test (enc/dec) (cdc/ede)