short-term TODOs:

   - code cleanup
   - logging what's worth logging, convert debug output into Log4Perl
   - generating jop id# automatically
   - cleanup strategy of memcached data after jobs are done
   - compute statistics how well slices are balanced out

midterm TODOs:

   - pipelined processor should not retrieve intermediary results
   - borrow more ideas from existing MR implementations (Perl, Python, ...)
   - constructor parameters to test liveness of servers
   - monitoring thread to extract status information as web service
   - web interface to look at monitoring info
   - increase of robustness against failures of
          - workers
          - memcacheds
     also during one run (revive SSH'ed workers)

   - support for testing algorithms
   - fake memcached to allow automatic smoke testing 

long-term TODOs:

   - build an application library under Parallel::MapReduce::Apps::*
   