Re: [boost] Re: Re: policy-based smart pointers are EVIL! (was: Intrusive smart pointer)

From: Dietmar Kuehl (dietmar_kuehl_at_[hidden])
Date: 2002-05-05 10:52:34

Peter Dimov wrote:

> From: "Dietmar Kuehl" <dietmar_kuehl_at_[hidden]>
>>David B. Held wrote:
>>>it seems that the only way for a policy-based pointer to survive
>>>would be for it to work with shared_ptr.
>>
>>This view does not reflect the discussion of the LWG and there were also
>>no votes or things like this. It was more or less a debate on where to
>>go and trying to find out the direction to move to. That is, if there is
>>a good system of smart pointers, shared_ptr would be abandoned: It is
>>not a fixed point the LWG has commited to use.
> 
> shared_ptr the name, or shared_ptr the semantics, would be abandoned? Or
> both?

Well, of course, this depends on what the result effectively is. I
actually meant to write a condition on this statement which I forgot:
If there is a good system of smart pointers which asks for abandoning
shared_ptr, shared_ptr would would be apandoned. Basically, I just
wanted to point out that shared_ptr is not [yet] set in stone and the
LWG does not feel bound to add shared_ptr to the standard or the TR.
I basically wanted to clear up that the statement of David Held was too
strong.

I hadn't looked into shared_ptr for a rather long time and from what I'm
reading here it seems to me that shared_ptr has evolved into a good
default for the smart pointer to be used in interfaces. Thus, I doubt
that the semantics would not be present in one form or the other if a
smart pointer makes it into the standard. Whether the name stays is a
different question but then, Germans used to say "Namen sind Schall und
Rauch" (names are sound and smoke; meaning that names don't really
matter - the named thing is what matters; however, this idiom is older
than programming languages and it has shown that names matter at least
to a certain extend).
-- 
<mailto:dietmar_kuehl_at_[hidden]> <http://www.dietmar-kuehl.de/>
Phaidros eaSE - Easy Software Engineering: <http://www.phaidros.com/>