Namburu, Clarke and Fox Conversations May 1 2001 -----------

A) In general DICE has become ICE -- Infrastructure to support Interdisciplinary Computation
	Gateway and DICE are largely complementary

B) Gateway has a distributed object model (CORBA), "seamless job submission", Kerberos
   security, graphical interface (old) for composing jobs
   
c) ICE has Separate Data Model and Data Format on top of HDF5
	These are better than XSIL (from Caltech) used in Gateway
	ICE allows Filters to transform data -- currently these are Scripts
	Gateway would call filters objects and view ICE version of filters as "inline object"
		natural to allow general objects as filters to allow more flexible codes
	ICE has powerful visualization

D) Some areas of crosscutting interest
	Object Model. Distributed versus Single Memory -- here Microsoft .net relevant
	Parameter Specification (as in  Ansys)

E) ICE concentrates on "Surface coupling" which implies a loose coupling between components
	This is same model as Gateway (linking of full applications as opposed to intertwining
		of separate application)
	Intertwined applications (called "Volume coupling") are harder as require modifying codes
		and high inter-application bandwidth (and low latency)
	DoE (Sierra from Sandia) and Common Component activities are targetting the hard "Volume
		coupling" case. They are not obviously succesful
	
F) What is relation to CCM Metacomputing activity we heard about from Lushington
	ICE would be called a "Grid Computing Environment" by Grid Forum

G) ARL is looking at methods from Graphics community to match meshes 
	Work with Haupt and Soni from MSU for meshes
	Chandra vision was a portal supporting "6 penetration codes with common mesh and visualization"
		This is a good goal

H) Fox invited Clarke to write paper on ICE for Grid Computing Environment Special Issue of
	Concurrency and Computation: Practice and Experience
		Due date is mid June 2001

I) We suggested collaboration of ICE with those outside ARL could involve:
	Addition of features as in linking with Gateway and as described in arlswoma.doc
	Use by other application groups to verify data model and data format are generally applicable

J) Need an ICE implementation that can be used by outside scientists
	Not clear if ICE can be taken outside DoD HPCMO
	Fox gave usual arguments why DoD would benefit from outside users
		ICE will be very hard to sustain unless ideas become part of community standards and practice