G'day all.
Quoting Conor McBride
How depressing!
Sorry, I don't understand that. Quotient types are good, but they're not the whole story. They just happen to be one use case with a solid history behind them.
it's just that we need to manage information hiding properly, which is perhaps not such a tall order.
It's my opinion (and I know I'm not alone in this) that modularity is probably the one thing that Haskell really hasn't (yet) gotten right. Haskell's implementation of modules/namespaces/whatever is the bare minimum needed to be minimally useful. It's a shame, because abstraction, in Haskell, is extremely cheap. It's often only one line, and you've got a compiler-checked abstraction that can't be accidentally confused with its representation. This should encourage micro-abstractions everywhere, but without submodules, namespaces or whatever you want to call them, these abstractions are easy to break (on purpose, not by accident). If only you could add a couple more lines of code, and instantly have your abstraction unbreakable. Cheers, Andrew Bromage