Comment by jl6

I think an argument can be made that relations, schemas and constraints encode a kind of business logic that is intrinsic to the definition and integrity of the data, while other types of business logic represent processes that may hinge on data but aren’t as tightly coupled to it. Similar to the difference between a primitive type and a function.

I guess some will argue that their business logic is special and really is so tightly coupled to the data definition that it belongs in the database, and I’m not going to claim those use cases don’t exist, but I’ve seen over-coupling far more often than under-coupling.

This is why I say: Applications come and go, but data is forever.