@warz, I was about to say the same, that solution would be completely unscalable.
I saw @drehb that you mention MD capacity in your post from way back.
I think my topics on indexed and searchable dbs can add some information. Like this post DataStore over AppendableData design - #20 by oetyng and following contents.
Or the OP in that topic.
There’s also a repo for the SAFE.DataStore that has rudimentary indexing and searching.
However… SQL-like querying of data will be available with SPARQL over the RDF-emulations that MaidSafe implement.
I am not well versed in the RDF world, but it would still seem to me that it maybe is not the silver bullet for storage (since basically no db type is, all dbs suck - at some stuff), and eventually we might want to some other kind of emulation/implementation as well.
So, as for building up indexes, that would be the first best bet anyway; available in the core, supports SQL like queries.