Usage:

  ./dayinlife <name> <n>

 name: some unique identifier, to avoid name space clashes 
       between instances running at the same time.

 n   : giving the number of iterations.  NOTE that, due to
       a resource leak in the underlying default job adaptor, 
       n > 49 will cause an bus error at the moment.


Details:

  The application consists of three parts: the main routine, the
  migration framework, and the workload.  For convenience,
  a logging facility is provided, too.


  The main routine is simple enough to show it here:

    {
      logger    l  ("dayinlife.log");
      framework fw (l, argc, argv);
      workload  wl (l, argc, argv);

      fw.run (wl);
    }

  A logger utility is created and shared between the framework
  and worker instance.  Both framework and worker evaluate
  argc/argv (the framework picks the name argument, the worker 
  the number of iterations).

  The framework does the following on run():

    - fetching input data to the current working directory
    - informing the worker about input and output data location
    - invoking the workloads work() routine once
    - invoking the workloads test() routine once
    - if that test succeeds, 
      - we are done, and finish
    - else
      - output data are collected
      - the whole application is migrated to
        a new host, and starts again.

  Note that the output data are registered in a logical file,
  which in turn is stored in the advert service.  That way, the
  consecutive instances of the application can communicate the
  data location to each other.  The output data are then used as
  input data for the new instance.

  The worker is in this case very simple: it reads a counter
  from input data, increments it, and writes it back to output
  data.  If the counter exceeds some maximum, nothing is done,
  and the test routine fails (nothing more to do).