Re: C++ PATCH: PR 20599 (1/3)

[Theodore Papadopoulo <Theodore dot Papadopoulo at sophia dot inria dot fr>]

On Tue, 2006-09-19 at 09:31 -0700, Joe Buck wrote:

> > Another way of giving the code some test is to find a way to enable it
> > only for the purpose of libstdc++.
> 
> I *disagree* with this.  GCC should be available as a testbed for features
> that are being seriously considered for future standards.  If users are
> prevented from using them, then it can't serve that purpose.
> 
> If features are needed to write a good standard library, then the same
> features are going to be useful for others to write good code (unless
> the standard library itself gives access to the full power of the
> feature).

Funnily, I also agree with this.
My mail was about to find middle-term positions that can be accepted by
both sides...

As I said, as a user, I'm eager to use variadic templates. On the other
hand, if C++ developpers believe that the feature is too "young" to be
put in my hands (which I would overcome certainly using branches,
patches or other means if I'm really motivated), having the mecanism
allowed for the library (simplifying the life of library developpers) is
already improving the users' life, give some "real life" exposure to the
feature which can be refined and improved and then exposed to more
people.

Then I guess that there might be some compiler magic that really helps
library writers, but are difficult to be made clean enough, general
enough, simple enough,.... for standard users. I do not think that
exposing users to such features is a good thing. This is not the case of
variadic templates, IMHO.