Or you could name the indexes something human readable, and import the
other context sets where appropriate. EG marc.245 is equivalent to
I personally wouldn't make the subfields a relation modifier, as the
content of the indexes would be very different.
It's similar to the way we wouldn't say:
dc.record any/element=dc.title "words in title"
dc.title any "words in title"
Of course you end up with a LOT of indexes if you enumerate every single
combination of field and subfield, but if there's a use case for that,
then so be it.
>So a query might look like
>marc.245 =/marc.subfield="a" "Smith"