Worry in the following order :

  Behavior

  API boundaries

  Data structures

  Implementation (flowchart) details


--== TODO ==--

review documentation

regenerate readme ( perldoc )

upload new version

cap the size of the file silos at 2G and split them further, so there is a 2.1, 2.2, etc silo that
is algorithmically calculated based on silo size and ID

investigate the possibility of having the record store spam multiple hosts and filesystems.

