- code goes into /src

- prose in /doc

- /experimental is a playground

- Name interfaces in the canonical Golang way.

- Skip the complicated logging madness for now. I do want
  dedicated logging for each job, but not right now...

- This now works just fine.

- Next up is (1) refactoring, (2) cleanup of the received ULO/RDF
  files and (3) better state information (for the web interface)

We now run jobs w/o limits. This is probably bad in a hosted setup but
we can worry about that later should it ever become a problem. As it
stands the implementation is easier with a simple goroutine rather than
a complicated pool of threads.