It would be convenient for solving a number of issues related to file versioning if we can break the link between the ECL code and the DFS information about who holds the master information about how a file is laid out on disk. At present they have to both say the same thing, and you might (if lucky) get an error at runtime if they don't match.
If eclcc connects to dali, we can improve this:
1. You can project out fields that are present on disk but are not listed in the ECL definition
2. You can project default values for fields that are present in the ECL but not on the disk file
3. You can use information about record counts to drive optimizations
4. Even if you don't want to project the fields, you can give errors (or warnings) of mismatches at compile time