Re: C++'s new 'export' Keyword?

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Bonzini <bonzini@gnu.org> writes:

| > :-) I would rather see a bit faster C++ compiler than implemented export. 
| > Is this feature so usefull? 
|  
| It is not, I don't give a damn about it. But Gaby is saying that the
| paper is bad and was debunked, 

    Message-ID: <40103296.EA992FE2@acm.org>
    Message-ID: <4010379D.253C1C85@acm.org>
    http://www.comeaucomputing.com/iso/promises.html

| but not *how the committee's view has
|  changed*. If the debunking was so important, I'd expect that he
| told us why export is so great. 

If I were you, I would try to  use it in real life and get my own idea
about it.  Most of the people who have done so found it useful for
their  purposes and were not "surprised" or "deceived" on any
particular expectations.  Jean-Marc Bouguet  accounted for personal
experience with "export" in the Overload journal (I think it is the
April 2003 issue).  Dinkumware has done so and they have made public
their experience.  If you think they are having illusions, then try it
by yourself.   
(I got a copy of it and tried to make an opinion by myself).  Just do it.

You probably should know that when export was voted into the C++
standard, the EDG guys voted against it. But they are also the first
person to implement it.  And they opposed to its removal on technical
grounds.  As Steve Adamczyk recalled at the Oxford meeting, when they
opposed to export adoption, they did not so because it was a useless
feature (they think it is a useful feature); they opposed because they
thought they did not know whether it would be implementable in a very
useful way.  They have found that they were mistaken; but then they
learnt a lesseon they are most willing to share.  They have shipped
their first implementation and Dinkumware has been shipping a standard
library that takes advantage of it. At last report, they said their
customers were happy with it. But instead of asking me, just go ask
EDG guys and Dinkumware.  They do have concrete experience with the
thing. 

Someone said that he does not have time to spend two years as it
happened at EDG.  That someone should probably contact the EDG guys to 
have an idea of how they spent time on implementing export.  
[ That someone might be surprised. ]

Anyway, all these have been rehashed at several times on C++ forums,
you should google a bit and read the discussions.

Someone said earlier in this thread that the papers in the pre- and
post mailings on WG21 usually have high quality.  Yes, but that is a
general comment.  And he built on that for saying that the paper
could not have contained inaccuirate materials.  That person is mistaken.
There is no "reviewing board" for WG21 mailing papers, and WG21 is not
a publishing journal.  The WG21 effort is an open and democratic
process.  If you think you have an issue and you can't explain it in
short and simple sentences, you've got to write up a paper and submit
it for the mailing and go to the meeting (or have someone go there)
explaining your issue.  The mere fact that a paper is part of the
mailing is *no endorsement* that the committee agrees with ideas 
expressed therein.  Papers that express the committee's ideas are
published either as a standard or technical reports.

[...]

| CodeSourcery. If one of Red Hat or CodeSourcery's clients believed
| that export is important, we'd already have an implementation of it. 

Yes, the usual self-fulfilling prophety.

-- Gaby