- man page
- lots of testing of new features, most notably -DLARGEFILES
- add rewrite test 
- add option to select which tests you want to run with logic
  which are the essential ones to run also to reach. ie if you want
  only seek test, writing files should still be made before them.
- add option to choose how you want to output of seek testing to appear.
  I still don't know which one is better MB/s or seeks/sec, perhaps
  even average seek time in milliseconds would be good. (miku)
- get the core perl wrapper of tiotest into a perl module and then
  various frontend perl modules can be interchanged for various
  kinds of output, even including something that does a use CGI,
  for instance :)
- figure out some nice easy standard for serializing the data, ideally
  to a simple enough xml file, so we can store data and re-analyze later
  with different front-ends.  xml serialization should ideally just be
  given the hash and bang it out - shouldn't need updating as the hash
  changes.
- the stat_data hash should be pretty solid - its format shouldn't need
  to change much, so our xml dtd would be fairly stable.
