Leo White

You wouldn't need to keep the information in signatures. Once the type has finished being defined the information is no longer needed -- you can just expand the type away...

It's not more modular because, as I said, the information cannot be usefully abstracted, and it's not more efficient because you need to do the expansion anyway to find out...

I agree that a binding operator is no worse than using `@@` with `with_foo` functions. However, it suffers from the same obvious drawback: it takes a construct whose entire purpose...

> Do you have examples in mind where the end of the scope induced by "let..in" is hard to determine? I don't think there is anywhere where it is genuinely...

I guess that we could use that. It seems a bit of a shame to add multiple `SUFFIX` directives to every .merlin file we make.

I think this is not a good idea. It will be confusing to users to get a type unequal to the one they asked for. The reason it works for...

Having it be an error would be fine. Although it won't actually stop you from making modules with such signatures, which I don't think can actually be achieved. For example,...

Sorry, I tried to shorten the example and it stopped being broken. Here is a version that actually has an alias to a functor parameter: ```ocaml # module type Type...

> it's pretty clear we want to keep the semantics of with module I would say that we want to strengthen the semantics, so that it adds a module alias....

> But Undefined is only an approximation of self-reference detection. It doesn't work for systhreads, or in fact any other domain-local concurrency mechanism: any cooperative-concurrency monad (or effect-handler) could lead...