Re: [boost] Koenig lookup, names in std:: - Summary

From: Matthew Austern (austern_at_[hidden])
Date: 2001-01-29 16:24:33

Peter Dimov wrote:

> > Well, before I get qualified a "biggot", I would like to make it
> > clear that I'm interessed in discussing formal proposal, not just
> > statement that something not clearly defined is a pure extension.
> 
> Then discuss the, more or less, formal proposal (LWG Issue 226) or provide a
> counter-proposal of your own.

I should probably remove that proposal from the "proposed resolution"
section of issue 226, because I think there's little chance that it
will be adopted.

That's an inflammatory statement, so now let me back off and clarify.
I do *not* say that partial specialization of function templates is 
unthinkable.  At the Toronto meeting, core was pretty explicit about
three things: (1) They are reluctant to add partial specialization of
function tempaltes.  (2) They realize that LWG 226 is a serious issue, 
and they will consider adding partial specialization of function 
templates if that's the only good way to resolve LWG 226.  (3) If they
do add partial specialization of function templates, they will want to
come up with their own proposal for doing it.  Nobody on core had any
killer objections to Peter's proposal, but they weren't completely happy
with the details of the syntax.

So that's where it stands now.  Core would rather see a library solution.
I think that most people on the LWG would also like to see a library
solution.  If LWG comes up with a convincing argument that there is no
good library solution, then core will think about how to add partial 
specialization of function templates.

My feeling is that at present, the leading candidate for a library solution
would be relaxing the restriction on user overloads within namespace std.
All other library solutions that I've heard of seem too much like hacks.

                        --Matt