If I understood what I just read in this article and on the MonetDB website, I believe if you were to implement the same logical design it would actually be all

If I understood what I just read in this article and on the MonetDB website, I believe if you were to implement the same logical design it would actually be all tables having two columns: one for the object id (OID, or row ID) and one for the particular data attribute. Then to build a tuple it would be a join on OID for the attributes you need. The physical design of the column oriented DB can omit the explicit OID because it is just the array position.