Ok, so the 2ndary indexing is still broken for Community edition... (but is supposed to be fixed in next release (6.1) )
The issue is that only JSON tables are capable of being indexed and the index uses the JSON tables and not a binary table. Because the Index is decoupled thru the use of a MapR Stream and stored separately from the other tables, it could be stored in a binary table or even in something outside of MapR.
The binary table would be faster as an inverted table than the JSON table.
Its also my understanding that since MapR controls the access to the indexing, they would also control how the index is being used and can control the APIs/interfaces to make this happen...
And along the same lines... abstract the read/write to MapR-DB so that the same API call can write to either Binary or JSON tables. (Which should actually be a different idea...)