"Salvatore Insalaco"
This doesn't look like a relational structure at all in Haskell.
I believe you are abusing terminology here. 'Relation' refers to a table (since it represents a subset of AxBxC.., i.e. a relation), not to references between tables. Mutability and mutability of references is of course important in most relational databases, but I'm not convinced an immutable database wouldn't be interesting and useful in a functional programming language. I've always (well not that I use them often) been annoyed at RDBMS lack of discriminated unions. The TH based approach by HAppS looks cool, but I think simply a slightly more general Data.Map (supporting multiple indices, search by named field and so on) could be a useful thing. -k -- If I haven't seen further, it is by standing in the footprints of giants