This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

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

On Jan 27, 2004, at 1:57 PM, Paolo Carlini wrote:

Gabriel Dos Reis wrote:

I guess it depends on whether you think "highly inaccuarate"[1] material
is a "good, free, introduction".

[1] Those are words used to describe that paper

Used by _you_? Why don't you explain a bit why exactly you believe those
papers are so bad? Otherwise, why Olivier (or me) should buy a whole book,
or the highly technical C++ Standard or bother Adamczyk himself only to
learn something about "Export"? "Export" is not mine or Olivier profession
and we don't plan to devote a couple of years to implement it, as happened
at EDG.

The fundamental problem is that the paper is largely about what it's like
to implement export, and it was written by someone who did not have
implementation experience. A large part of what was in that paper was
literally hearsay (i.e. saying what someone else said, where in this case
the "someone else" was Daveed Vandevoorde of EDG). Instead of
asking Herb Sutter what Daveed Vandevoorde said, it's better to go and
ask Daveed himself---by which I mean either buy Daveed's book, or
literally ask him. He's a nice guy, and he'll be quite willing to answer
questions about export.

(The other part of the paper is a demonstration that it's possible to create
examples using export that behave in ways that a novice might find
surprising. Sadly, that part of the paper is right. Even more sadly, export
isn't the only language feature that could be useful in an Obfuscated C++
contest. My opinion is that export is not even close to being the worst
offender in this regard. There are lots and lots of ways that you can write
C++ torture cases that behave differently than someone might expect at first

But yes, if we choose to implement export then it will be a major task.
It probably won't be as hard for us as it was for EDG, partly because the
second implementation is always easier than the first and partly because
EDG has volunteered to help anyone who's planning to implement export.
(EDG wants to have the first and best export implementation, but they don't
want to have the only one.) No denying that it's a big project, though.
That's why nobody has started work on it yet.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]