Re: [boost] plain pointers and shared_ptr

From: Fernando Cacciola (fcacciola_at_[hidden])
Date: 2002-03-26 14:54:43

----- Original Message -----
From: "E. Gladyshev" <egladysh_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Tuesday, March 26, 2002 4:01 PM
Subject: RE: [boost] plain pointers and shared_ptr

>
> > > All that said, none of this really changes the
> > problem domain.  There simply
> > > isn't a universal solution for what the poster
> > wants.
> >
> > Agreed.
> >
>
> What do you mean, "there simply
> isn't a universal solution"?
> Some people on this forum are already
> solving this problem just try to be
> a bit more positive and open-minded,
> it is just a fun after all :).
>
>
No, the people in this forum have already solve other problems, but not
this.
Your problem is underspecified, and that's way there cannot possibly be a
solution.

The reason why it is underspecified is that you insist, and actually push,
on the 'know-nothing' parts -which is OK-, but without stating the known
parts at all. That leaves your problem undefined.

IOWs, every system, as a whole, must be completely specified, so each time
you take responsibility out of a piece of code, you are required to put it
in on some other piece, either by direct coding or by contract
(specification).

You keep telling us what foo() cannot know or do... OK, but then you must
specify who can do what exactly, otherwise the solution is unbounded.
I couldn't yet understand who and how is to be responsible of memory and
object lifetime management in your problem.
You need to settle this down one way or another *before* looking for a tool
to do it (such as a smart pointer or a garbage collector)

Fernando Cacciola
Sierra s.r.l.
fcacciola_at_[hidden]
www.gosierra.com