If you have a look at the literature, for example the proceedings of VLDB, you'll see that nested relational databases are, in general, the preferred theoretical model. There are a few quirks to iron out where the semantics don't quite align, but on the whole they are the better approach.
Why? Because the relational component still allows set-at-a-time interaction for efficient querying, but path expressions can also be used to navigate through the nested structure.
...or RMI. Really, if they want to change the transport layer from IIOP to somthing over HTTP then why not produce an IDL compiler that spits out stubs using their new wiz-bang protocol instead. Why introduce yet another IDL, yet another 'standard' set of naming & trading services, etc.?
Slashdot Mirror
Why? Because the relational component still allows set-at-a-time interaction for efficient querying, but path expressions can also be used to navigate through the nested structure.
...or RMI. Really, if they want to change the transport layer from IIOP to somthing over HTTP then why not produce an IDL compiler that spits out stubs using their new wiz-bang protocol instead. Why introduce yet another IDL, yet another 'standard' set of naming & trading services, etc.?