Vasilis Papavasileiou

@Drup, you can use the `Eliom_registration` API, or the more liberal `Unsafe` submodule. Or we can even provide an `Eliom_service.Ocaml`, no big deal, it is one function. Which partially brings...

In fact, no submodules are really needed, just a `create_unsafe` and a `create_ocaml` with different shadow parameters.

It is just two functions now, not 2 \* 12, and the duplicated ones will be aliasing `create` implementation-wise (I think), and under a `(**/**)` in the signature. So no...

There is `create_external` also which would need to be duplicated (I suspect that remote OCaml services can be useful). So merge `create_external` into `create` and do the above? Not sure.

I pushed a branch doing the above: https://github.com/ocsigen/eliom/compare/no-service-ret . It eliminates the need to explain `Ret.t` and the duplication is insignificant. I think I like the net effect.

That's strange. Could you test on a non-Flambda switch? It should work fine there. If the problem is related to Flambda, we would need to investigate what happens twice. I...

I can't reproduce. Please provide the complete code that you are running.

Your `Eliom_registration.Html_text.create` call registers the service automatically, and your `Mysite_app.register` call tries to register it a second time. You were supposed to replace the contents of `mysite.eliom` with the provided...

@bluddy, we need to rephrase the _instead of the default one_ part, I understand how one would get confused. @Drup, it sounds feasible. We could store the registration locations (obtained...

There is an exception already, and we could show a backtrace during the second register. I don't know how much that would help new users; they would have to spot...