* low level routine for NGAS direct copy download
* low level routing for NGAS streaming download

* get a file by NGAS id:
    * routine that figures out which server the file is on and
      calls the low level routines

* get a file by AAT file id:
    * routine that figures out what the NGAS id is and calls
      the 'get a file by NGAS id' routine, and renames the
      result to match what the AAT db says it is

* get a fileset by fileset name:
    * routine that builds a list of files in the fileset, walks
      through them and blah

I need a standard way to describe arguments to the low level
routines and the ability to generate these arguments based on
what the user wants.

source name
source location (path, NGAS server, whatever)
destination name
destination directory
expected file size

generalized steps for a datafetcher request:
    * look up the missing metadata (filename, ngas_id, whatever)
    * decompose the request into one or more parts
    * get all the parts
        * in parallel if possible & more than one