John G's thoughts about Jim's mail:

I have a question about the data collocation discussion.  For a single project or site, I agree: the cost-benefit ratio of a distributed system is lousy.  But if your data is in fact collected at different sites, or by different organizations, or by different projects with different funding sources and different requirements -- in those cases isn't distributed data a fact of life?  Or is it also true that in those cases -- say for example with IOOS -- that there should be just one big data repository (or a bunch of identically organized little repositories) managed by a central authority?

Here are some "success stories" to consider, where the data is distributed in different organizations:
 - Any web search engine
 - Many business enterprises (via the web services protocols like SOAP, WSDL, and UDDI)
 - People looking for oceanographic data (e.g., using the OPeNDAP THREDDS server architecture)
 - People looking for GIS-based data content (using WorldWind, for example, or OGC protocols)
None are perfect solutions, but all have made great strides, because they defined protocols -- or took advantage of existing protocols -- that allowed all those different sources of data to be combined.

Distributed data access can be pretty straightforward, IF everyone agrees on the protocols for the exchange of the information.  In fact, if the data you want to use is collected at different physical locations, then by definition you have to use a distributed data system.  The real challenge comes when different groups manage their data sets in different ways, which is the case for any data that isn't "ours" and much of the data that is "ours". That's why I think it's important to create the best way to get those systems to interoperate, so that they can be interconnected as much as possible.

Many enterprises have managed distributed,  heterogeneous data at least somewhat successfully, and at relatively low cost to the sharing institutions.  It's just a matter of finding the right combination of technology and social engineering.  Who better to do that than MBARI and some of its closest friends?