Re: RFC: mp = &o.f;

[Joe Buck <jbuck at synopsys dot com>]

On Mon, Mar 17, 2003 at 04:56:17PM +0100, Christian Parpart wrote:
> However, that, what I am talking about was a proposal by Borland already made
> in september 2002.
> 
> The refering URL is:
> http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/papers/2002/n1384.pdf

That document makes three proposals.  The first one, bound pointers to
members, I would heartily support, as it would greatly simplify a lot
of the code that I (and others) write.  Practically every approach to
simulation in C++ has to simulate this concept of bound pointers somehow
(either with a common baseclass or with templates, and both means have
their disadvantages).

I'm less enthusiatic about the others (properties and extended RTTI).

It might make sense for GCC to accept clean implementations of serious
standard-extension proposals, but we'd need to be very cautious about
it.  If we accept everything, it hampers maintainance; if we accept
nothing, it puts GCC behind the competition (after all, standards
committees want to accept proposals based on demonstrated usefulness,
so implementations must precede standardization).