Index harvesting vs federated query

[dennisvang]

Currently the FDP-index "harvests" all metadata from the registered FDPs (using the LDP navigation links), in order to be able to run SPARQL queries. In other words, the index copies the metadata from all FDPs into its own (local) triple store (a.k.a. data centralization?).

I'm wondering why the index does not use some kind of federated query approach. Basically the index would send the query to all registered FDPs (via their API), to let them execute the query on their own triple stores, after which the index would join or aggregate the results in some way.

There's probably a good reason for this...?