Seems like we are adding more and more options and commands that take
a file or URL argument, where the file can be one of a number of
places. Perhaps we can unify things a bit by agreeing to use URL
conventions in these places?
file:// is a file on the current machine
http is supported everywhere
hdfs://host[:port]/ is the used to refer to HDFS files
hdfs:/// is used to use the default hdfs?
rsync, ftp and https are supported too (eventually)
torrent:// is used to support bittorent someday...
Should we standardize on such a conventions and start digging to put
a class together to make this all seamless?