Discussion about Kepler proposal

• "Technology-enabled science" --- really good idea, but rest of presentation's screenshots don't really support/follow through on this concept.
• Lots of overlap with (early) DCS concepts; almost like they read our material!
• We should focus more on describing the data set, rather than building too many small tools for different projects. Then more people can deal with it in their analysis themselves.
• Functional MRI Analysis slide is really good: illustrates "tpyed data channels", nice high level names for their data types.
• Their tools imply that the modules themselves name their input types. This is a good thing; it saves the user from having to know/remember what can talk to what.
• Workflow is fixed for some applications, but investigative or research isn't a fixed workflow at all; it's about developing new ideas. Therefore, it's key that the end user has lots of flexibility about what they can work with, and how they're working with it. Contrast: Oracle vs. (old) Delphi
• "Grid enabled" where "grid" references "grid computing"

Main question: should we join Kepler?

• We could definitely use some of their stuff, easier than starting over.
• But how much? How much will we want to use, how much will we have our own solutions for.
• Making the web searchable for certain kinds of data
  • Very compatible with Kepler
  • But not overlapping; they haven't really dealt with it yet
  • They're dealing with "once you've already got the data" problem.
  • Our crawler idea might be a good complement.
• Makes sense to start farther ahead than we'd otherwise be by starting from scratch.

For Pat:

• Make some inquiries with these guys.
• Go as far as you want; feel free to drop names and represent us!

-- BobKrzaczek - 28 Mar 2005