Re: [development] Referential integrity -- finally?

Dries When it comes to abstraction, one has to differentiate between two things. When it comes to DML (Data manipulation language, such as INSERT, SELECT, UPDATE, ...etc.) abstraction normally becomes a lowest common denominator and performance may suffer. For DDL (Data definition language, such as CREATE), there is no standard for column types and syntax, and hence abstraction buys us portability. Sites that need tuning (e.g. adding an index, ..etc.) will do so regardless of whether we do abstraction or not. Therefore, I am opposed to too much abstraction in DML, but happy with abstraction, as suggested by Adrian in the issue mentioned, for DDL.