Frankly, understanding of what the relational model actually is has been missing from the industry even among many people who have made data management their primary focus. IMHO a lot of the penetration of the NoSQL stuff in the last decade was based on misunderstandings of the capabilities and potentials of the relational model.
I think the early history of SQL was too influenced by ISAM-type stores and COBOL flatfiles and the like. Hence the vocabulary around "tables" and "columns" instead of the relations, relvars, tuples, etc. described by the actual relational algebraic model. To many of those with an OO fetish, "relational model" became synonymous with "tabular model", and the industry's early 2000's obsession with object polymorphism, OORMs middleware, etc continued the stall in relational database innovation.
I feel like there is now renewed interest in this stuff, though. Your project looks really neat, for example.
I think you analysis is correct. I also add that is related too to the languages, you point with Cobol, and I also add C, Pascal, Java, etc.
When they have nulls and limited to zero queries, relationships and data modeling capabilities is not hard to see that how they are is reflected in how the DBs are too. I honestly not know about sum types until a few years ago, neither have a clue how much useful they are, so surely even today many are in the same blind spots...
I think the early history of SQL was too influenced by ISAM-type stores and COBOL flatfiles and the like. Hence the vocabulary around "tables" and "columns" instead of the relations, relvars, tuples, etc. described by the actual relational algebraic model. To many of those with an OO fetish, "relational model" became synonymous with "tabular model", and the industry's early 2000's obsession with object polymorphism, OORMs middleware, etc continued the stall in relational database innovation.
I feel like there is now renewed interest in this stuff, though. Your project looks really neat, for example.