Tests that would be nice to have

OR TESTS
========
* Initialize new OR and check directory structure --- Done
* Create objects and try fetching them back --- Done
* Modify objects, and then fetch back and check for new value
* Create some new object, crash OR, recover, check values
* Modify objects, crash, recover, check values
* Create lots of new objects to cause disk writes.
  Check values of objects that should have been flushed out of the log.
  Then crash, recover, and check values.
* Check that concurrent transaction commits retain consistency:
  - run multiple clients
  - each client transfers some amount of data from one object to another
  - run auditor that repeatedly checks that data values always add up
* Check that objects are accessible even while being written out:
  - set log size to small
  - start modifying set of objects in round-robin fashion
  - check object values often

CLIENT/FE TESTS
===============
* Reverting of objects after an abort --- Done
* Clients running in parallel generating lots of aborts
* Handling of exceptions in general (occuring at different points)
* Test object eqyality for basc values and normal objects

* Test the directory object --- Done

IMPORTANT:

* After a transaction abort --- handle the exception and check that there
  are no objects marked as read/written in this new transaction
  Try this at different points	

