From mboxrd@z Thu Jan  1 00:00:00 1970
Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id OAA24731 for caml-redistribution; Fri, 17 Sep 1999 14:13:01 +0200 (MET DST)
Received: from concorde.inria.fr (concorde.inria.fr [192.93.2.39]) by pauillac.inria.fr (8.7.6/8.7.3) with ESMTP id SAA32204 for <caml-list@pauillac.inria.fr>; Thu, 16 Sep 1999 18:59:42 +0200 (MET DST)
Received: from pauillac.inria.fr (pauillac.inria.fr [128.93.11.35])
	by concorde.inria.fr (8.8.7/8.8.7) with ESMTP id SAA10694;
	Thu, 16 Sep 1999 18:59:39 +0200 (MET DST)
Received: (from xleroy@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id SAA32366; Thu, 16 Sep 1999 18:59:39 +0200 (MET DST)
Message-ID: <19990916185939.24663@pauillac.inria.fr>
Date: Thu, 16 Sep 1999 18:59:39 +0200
From: Xavier Leroy <Xavier.Leroy@inria.fr>
To: Francois Pottier <Francois.Pottier@inria.fr>,
        "Pierre CREGUT - FT . BD/CNET/DTL/MSV" <pierre.cregut@cnet.francetelecom.fr>,
        caml-list@inria.fr
Cc: caml-list@inria.fr
Subject: Re: Parameterized signatures needed ?
References: <19990913122532.A4032@lsun605> <19990914114600.57679@pauillac.inria.fr>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
X-Mailer: Mutt 0.89.1
In-Reply-To: <19990914114600.57679@pauillac.inria.fr>; from Francois Pottier on Tue, Sep 14, 1999 at 11:46:00AM +0200
Sender: weis

> In view of the
> solutions which exist at the level of expressions, one may suggest
> extending O'Caml with
> 
>  * either explicit sub-signature constraints, e.g.
> 
>    module d(X:sig
> 		 module type B < sig type t end
> 		 module F(a:A) : sig
> 		     module b:B
> 		     module c:C with type t = b.t
> 		   end
> 	       end) : ...

It looks like this would fit quite naturally within the OCaml module
system.  After all, the module language is explicitely typed, has
subtyping and a form of parametric polymorphism; bounded polymorphism
is the next step.  (I haven't checked the details, though.)

>  * or row variables in signatures, although I am not sure which form
>    this would take.

This looks less natural to me.  A very limited form of "row
polymorphism" can already be expressed with module type parameters and
sub-structures, as in:

    module d(X:sig
                 module type X
 		 module F(a:A) : sig
 		     module b: sig type t  module M: X end
 		     module c:C with type t = b.t
 		   end
 	       end) : ...

but it's very limited and not really practical.

- Xavier Leroy