The row orientation of a sql database is a significant limitation for sufficiently deep datastructures IMO. They often force you into multiple calls for a single entity or complicated loops through really long rows. All of which can be the cause of wasted cycles in code. A sql db with its rows presents a poor datamodel for those cases.
You are confusing SQL with "necessitates row ortientation", even tho column databases frequently use SQL for queries. This is essentially why I originally asked -- SQL and relational data does not require what you describe, it has just been the traditional approach to implementing it.
I think when zaphar said "sql database" he was just using common usage for RDBMS. Yes, "SQL database" does not have to mean "row oriented storage". But all of the commonly familiar SQL RDBMSs use row oriented storage.
I actually think SQL the language assumes row oriented storage. I honestly can't think of a single SQL database that isn't row oriented. While theoretically you could, in practice no one has.
I can't say for certain but this might be related to the reliance on sql for the query language.
