Re: [Caml-list] Are record types generative?

[Jeremy Yallop <yallop@gmail.com>]

On 23 January 2018 at 21:36, Chet Murthy <murthy.chet@gmail.com> wrote:
> Heh. I was careful to say "caml", not "ocaml".   I'm aware that polymorphic
> variants, objects, and (heh, I forgot) labels all violate the rule. And of
> course, modules do.
>
> Like I said: *caml* has this property.  Certainly caml-light, circa 1993
> did, IIRC.

I think people usually consider OCaml an implementation of Caml.
(See, e.g. the documentation at http://caml.inria.fr/)

Of course, you're quite right that most of these programs weren't
valid in caml light.  But even (later versions of) caml light had
format strings with string syntax, so it was possible to write
programs that were rejected without annotations and accepted with
annotations.

In any case, whether annotations can be erased without affecting type
correctness is an interesting property to consider.  Here's a closely
related property that's much harder to break: does adding annotations
leave the run-time behaviour of a program unchanged?  There are far
fewer programs that violate that property, I think.