From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from mail4-relais-sop.national.inria.fr (mail4-relais-sop.national.inria.fr [192.134.164.105])
	by walapai.inria.fr (8.13.6/8.13.6) with ESMTP id p7IAgTxb014961
	for <caml-list@sympa-roc.inria.fr>; Thu, 18 Aug 2011 12:42:30 +0200
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AnABACHsTE7B/BfVlGdsb2JhbABCqQMBAQEBCQsJCRQDIoFAAQEFOEARCxgJFg8JAwIBAgFFBg0IAQGHbrkShkgEkxOFDItr
X-IronPort-AV: E=Sophos;i="4.68,244,1312149600"; 
   d="scan'208";a="105748314"
Received: from msa04.smtpout.orange.fr (HELO msa.smtpout.orange.fr) ([193.252.23.213])
  by mail4-smtp-sop.national.inria.fr with ESMTP; 18 Aug 2011 12:42:30 +0200
Received: from [192.168.1.63] ([83.199.49.118])
	by mwinf5d66 with ME
	id Mmio1h0032Z0S2z03mioqx; Thu, 18 Aug 2011 12:42:49 +0200
X-ME-engine: default
Message-ID: <4E4CEC97.9010202@frisch.fr>
Date: Thu, 18 Aug 2011 12:42:31 +0200
From: Alain Frisch <alain@frisch.fr>
User-Agent: Mozilla/5.0 (X11; U; Linux i686 (x86_64); en-US; rv:1.9.2.20) Gecko/20110804 Thunderbird/3.1.12
MIME-Version: 1.0
To: caml-list <caml-list@inria.fr>
References: <4E4CC071.1060900@inria.fr> <20110818080254.GA10854@ccellier.rd.securactive.lan> <CAPFanBHyuEpJ4NE61Rdvq2A8PZ4Uuk-inUCXJHO-yrZFDmVtxQ@mail.gmail.com>
In-Reply-To: <CAPFanBHyuEpJ4NE61Rdvq2A8PZ4Uuk-inUCXJHO-yrZFDmVtxQ@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [Caml-list] [Ann] Zarith

On 08/18/2011 10:58 AM, Gabriel Scherer wrote:
> I believe the comparison hardening was made desirable by the expected
> use of first-class modules to encode existential datatypes: with
> existential datatypes you may try to use the polymorphic comparison
> operators on two values of the same (existential) datatype having
> different internal representations. The problem didn't happen with the
> previous encoding of existential types using first-class polymorphism,
> as it implies packing the value inside closures, so the polymorphic
> comparison operators were not usable.

The problem existed before, because of exceptions. When two exception 
values having constructors with the same name are compared with the 
polymorphic equality, their arguments are compared structurally even if 
they don't have the same type.  (One way to address this would be to 
have a special tag on exception values to force physical comparison of 
the constructors.)


-- Alain