[boost] Re: Re: Re: TR2 and C++0x heads up

From: Beman Dawes (bdawes_at_[hidden])
Date: 2005-04-26 17:28:00

"Thorsten Ottosen" <nesotto_at_[hidden]> wrote in message 
news:d4lgqm$83h$1_at_sea.gmane.org...
> "Stefan Seefeld" <seefeld_at_[hidden]> wrote in message
> news:426E4259.4070209_at_sympatico.ca...
> | Gennadiy Rozental wrote:
> |
> | > IMO
> | > the more flexible the solution is - the more users it will end up
> | > attracting.
> |
> | I don't agree with this statement at all. Making things more flexible
> | incures a semantic dilution (not to speak of the added complexity
> | necessary to support that flexibility) that encourages users to just
> | 'roll their own' solution.
> |
> | I'm not saying that flexibility or genericity is bad. But it comes
> | with a price.
>
> I don't think it is unfair to see it as an expert tool and not something 
> the
> ordinary user need to use or worry about learning. That implies that we do
> want it in boost,  but that it might not be a good candidate for std::.

The plan we talked about several years ago was that if the standard library 
has a PBSP, there would also be classes (or similar typedef templates, 
depending on core language changes) like:

    template<class T> class shared_ptr : public smart_ptr<T, one set of 
policies> {...};
    template<class T> class scoped_ptr : public smart_ptr<T, another set of 
policies> {...};
    etc.

These would be recommended to non-expert users. One of my personal criteria 
for a good PBSP is that it can supply the feature set needed to provide 
those pre-configured smart pointers.

--Beman