TODO

TODO:

DONE:
  * Make it possible to clean up jobs through something like 'crowd cleanup'.
  * Think about adding a CLEANUP state, with WorkUnits being dispatched to
    workers to clean up a Job's leftover files on S3, instead of the current
    in-request cleanup. (see Job#cleanup_assets)
  * Right now actions just process. Think about map / process / reduce.
  * Basic unit tests.
  * Job creation UI. Pick an action, add inputs, set options, go. (not needed)
  * Look into Grand Central Dispatch for ideas about how to implement
    Job ordering/piping around.
  * Finish fleshing out the crowd command.
  * Security. HTTP Basic Auth via config.yml,
  * Security. S3 authenticated URLs.
  * Use the email field to send mail when the Job's gone bad. (bad idea)
  * Benchmarking mode -- benchmark every single aspect of a job and report it.
  * Pare down the code that the workers are loading to the absolute minimum.
  * Documentation. (document inside of config.yml, too)
  * Think about having the completion of a work unit respond with another one,
    instead of having to make a second request.
  * Think about having a failed work unit shuffle off to the end of the queue,
    instead of going again right away (and probably being handed back to the
    very same daemon).
  * Replace work_units_remaining with a percent_complete
  * Monitor UI -- see queue of jobs / work units.
  * Basic integration tests.
  * Unit tests for lib.
  * Bump to version 0.1 and release on RubyForge.
  * Re-architected to use real Nodes (version 0.2).