Galaxy has been suggested internally:
	Why not start with open source & build what's missing, instead of
	building all from scratch?

Galaxy has been tried:
	see Refinery project at HMS/Park Lab
	not perfect, either: lots of custom coding was needed
	had to go outside galaxy to handle data loads
	no workflow restarting from partially complete state
			only restart from beginning
	no workflows of workflows
	no job avoidance:  huge for TCGA ... data change each month, but
		only in drips & drabs
	significant manual interaction required for large #s of files
			(100s of files GBs in size)
		BUT we have 1000s of files, TBs in size
	freezing
	histories
	sample-based or file-based?
	scaling:  150K jobs in Galaxy 1 month (June 2012)
			  250K jobs in FH 1 week + >= 100K or more avoided >= ~350K jobs per week

How many million jobs have been run through synapse in 1 month?
Has it been tested on all diseases, all datatypes, all analyses for all AWGs?