[rstuart4133]

I don't know about not making sense from first principles, but if a principle concern is bug free code then NULL is a disaster. I've seen so bugs caused by people forgetting a column could be NULL and using '<', '<=' I avoid them like the plague now. It's not just the unfamiliar behaviour, it's also that everyone forgets to write tests for it.

A simple fix would be to add a dialect option that makes any operation on a potentially null value a syntax error. This is possible because in SQL it's almost always possible to determine if a value X could potentially be NULL or not. If it could be NULL, something like (X == y) or (X + y) should generate a syntax error. Instead you should have to write (X is not null and X == y) or ISNULL(X + y, 0).