ark@research.att.com (Andrew Koenig) writes:

|>  In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
|>  Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:
|>
|>  > Here's a suggestion for the next C++ standard: DON'T develop it
|>  > through the ISO process.  Instead, establish a separate non-profit
|>  > volunteer "free standards" organization and develop it through that.
|>  > Release the standard freely and distribute it on the internet.
|>
|>  We thought of that, actually, but it doesn't work--for entirely
|>  political reasons:  Any collection of vendors that got together
|>  to agree on a standard for something as important as a major
|>  programming language would instantly be sued for antitrust!

So how does OMG get away with it?  Is this in any way different from the
CORBA standard?

--
James Kanze    +33 (0)1 39 23 84 71    mailto: kanze@gabi-soft.fr
GABI Software, 22 rue Jacques-Lemercier, 78000 Versailles, France
Conseils en informatique orient   e objet --
              -- Beratung in objektorientierter Datenverarbeitung


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Stephen.Clamage@sun.com (Steve Clamage) writes:

|>  James.Kanze@dresdner-bank.com writes:
|>
|>  >The exact quote is "VisualAge C++ supports the C++ language and
|>  >libraries specified in the ANSI/ISO standard adopted in July 1998."
|>  >That's *not* the final standard.
|>
|>  Yes it is.

My mistake.  I was off by a year; for some reason, I was thinking that
the September 1997 meeting was in fact in September 1998.

--
James Kanze    +33 (0)1 39 23 84 71    mailto: kanze@gabi-soft.fr
GABI Software, 22 rue Jacques-Lemercier, 78000 Versailles, France
Conseils en informatique orient   e objet --
              -- Beratung in objektorientierter Datenverarbeitung
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Hi,

ark@research.att.com (Andrew Koenig) writes:

> In article <m390d0xjyh.fsf@gabi-soft.fr>, J. Kanze
> <kanze@gabi-soft.fr> wrote:
>
> > So how does OMG get away with it?
>
> I don't know.  Perhaps someone else does?

 As far as I know the OMG charges its members.  The OMG website
explains how to become a member as well as the fees and rights that
come with the membership.


  Just my two cents.

--
Carlos O'Ryan (coryan@cs.wustl.edu)
#include <std.disclaimer>
#include <cute.quote> // "Speak softly and carry a megawatt laser"
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <lyubthvd25r.fsf@macarena.cs.wustl.edu>,
Carlos O'Ryan <coryan@cs.wustl.edu> wrote:
> As far as I know the OMG charges its members.  The OMG website
>explains how to become a member as well as the fees and rights that
>come with the membership.

How does charging for membership prevent them from being charged with
anti-trust?  That's what was referred to by "get away with it".

I think the answer may be that the government has some special exceptions
for high-tech consortia, which were created to make it easier for the US
computer industry to compete with Japan.  That's how organizations like OSF
also get away with it.

--
Barry Margolin, barmar@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James.Kanze@dresdner-bank.com wrote:
>
> In article <36E5BB00.F70A0AEC@wizard.net>,
>   James Kuyper <kuyper@wizard.net> wrote:
> >
> > Martin von Loewis wrote:
> > >
> > > James.Kanze@dresdner-bank.com writes:
> > >
> > > > I have been told (indirectly) that there are at least two which are
> > > > shipping: Visual Age C++ 4.0 and Sun CC 5.0.  I've not been able to
> > > > verify personally whether either are "100% compliant"; maybe someone
> > > > from IBM or Sun would care to confirm.
> > >
> > There's no weasel-wording about their claims of standard compliance, but
> > also very little detail. That tends to make me suspicious.

[snip]

> > > Visual Workshop C++: Support for ANSI/ISO C++
> > >                      (ISO/IEC14882-1998 C++ language standard)
> > > (http://www.sun.com/workshop/visual/)
> >
> > That site merely says that it "Supports" the standard. It gives a lot
> > more detail, if you look into the white papers, and some of that detail
> > indicates incomplete compliance:
> >
> > "The C++ standard has some new rules for templates that make old code
> > nonconforming, particularly code involving the use of the new keyword
> > typename. The 5.0 compiler does not yet enforce these rules, but does
> > recognize this keyword."
>
> If the compiler supports all code conformant to the standard, and is
> just not enforcing new rules, no problem.  If it is just recognizing the
> new keyword, but doing nothing about it, however, "supports" is not in
> order.

The Sun WorkShop Compiler C++ 5.0 is our first implementation
supporting the majority of the features in the ISO C++ Standard.  For
various reasons, it does not support all of the features.  We have
documented the major features that we do not support in the "c++"
readme distributed with the compiler:

    The C++ compiler (CC) supports the ISO standard for C++,
    ISO IS 14882:1998, Programming Language C++. The following list
    describes requirements in the standard not supported in this
    release:

     o new syntax for function template calls: f<mytype>(args)
     o non-type template parameters for function templates
     o class member templates
     o partial specialization of templates
     o template template parameters (templates as template parameters)
     o universal character names

As we complete the C++ 5.0 FAQ, this information will be provided
there.

Our compiler team's goal is to have a solid implementation of the
standard. Part of the reason some of the features are not fully
implemented is that the team does not feel that some features are
adequately defined.  For example: the potential interactions between
standards compilation model and template template parameters.  Another
reason was simply release schedules and market demands for an update.

As you point out, this is a new compiler.  The "Standard Mode" of the
compiler is basically a 1.0 release of the front end. It has been
through an extensive Early Access program, but it still has a number
of bugs to be fixed.  We want to encourage users to use the compiler
and continue to file bug reports
(http://www.sun.com/workshop/technical/bugreport.html).  Also keep an
eye out for patches to the compiler.  We will probably be updating
patches every few months.

The 5.0 C++ compiler also implements a "Compatibility Mode" which is
binary and source compatible with our previous C++ 4.x compilers. The
compatibility mode is based on the ARM definition of the language.
This mode of the 5.0 compiler is very stable for the features that it
implements.

[snip]

> > Still, if that's the entirety of the non-compliance, it's not too bad!
> > The prices are, however, way out of my ballpark.
>
> N'est-ce pas?  Sun used to sell the compilers, unbundled, for a
> reasonable price.  Now, however, they only sell them bundled with a IDE
> and a debugger, at the top price, and I have problems getting an upgrade
> on my maintanance contract because the product I bought doesn't exist
> any more.
>

[snip]

Sun does still sell only the compilers.  However, those items are not
listed for sale on the web site.  They should be available through a
Sun sales rep or reseller.  The least expensive environments for C++
is either "Sun Visual WorkShop C++ Personal Edition" or "Sun WorkShop
Professional C" with "Sun WorkShop Compilers C/C++."

To obtain only the C++ compiler, the minimum environment is "Sun
WorkShop Compilers C/C++."  Please note that you do not receive a
license for dbx with this product bundle, so the functionality is
limited.

Ben Vizzier
Sun Microsystems Inc.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Ben Vizzier <ben.vizzier@eng.sun.com> wrote:
:     The C++ compiler (CC) supports the ISO standard for C++,
:     ISO IS 14882:1998, Programming Language C++. The following list
:     describes requirements in the standard not supported in this
:     release:

:      o new syntax for function template calls: f<mytype>(args)
:      o non-type template parameters for function templates
:      o class member templates
:      o partial specialization of templates
:      o template template parameters (templates as template parameters)
:      o universal character names

In one of my previous posts I implied that in my view SunPro 5.0
is not an ``archaic'' compiler and that it is not very clear-cut
whether the much-vaunted bazaar egcs or the much-vaunted non-bazaar
SunPro 5.0 did more in the last 5 years.

It was caused by my unfamiliarity with the limitations of SunPro 5.0.
I take it back on both counts.

Oleg.
--
Life is a sexually transmitted, 100% lethal disease.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

gjohnson@dream.season.com (Reality is a point of view) writes:
>>  +---- bs@research.att.com wrote (7 Mar 1999 17:35:14 GMT):
>>  | Now is the time to plan and experiment for the future, not to
>>  | quibble over the past.
>>  +----
>>
>> Any estimates on how many more months will pass before a
>> reasonably complete ISO C++ implementation appears?

Bjarne Stroustrup wrote:
> For some reasonable definition of "reasonably complete" this happened
> a few months ago.
>
> I can do most of what I need to do using EDG-based compilers, GNU's
> EGCS, MS and Borland compilers, and Sun just released their 5.0 which
> is claimed to be pretty close.

Some of us who write code that is ported to multiple platforms
(Unixes,  Win32, etc.) are not so lucky.  We expect full support
for exceptions, template specializations, template member functions,
and even basics like nested classes to appear for our commercial
target machines in about 12 months or so.  In the mean time, we
program with a subset of ISO C++ (which I dubbed "GCD-C++" a few
months ago).

Also remember that upgrading to a new version of a compiler is not
as simple as flipping a switch; makefiles need fixing, buggy code
needs fixing (some code that used to be valid is now illegal),
development machines require updating, etc.  And project manhours
need to be scheduled to get all of this done.

Then there's the problem of getting our customers to upgrade their
systems (so they can use our new libraries and header files).  This
is especially painful if you're in a multi-department company like
the one I work for, where each development group has its own
build environment.  Multiple products from the same company are
interrelated to each other as well as to the O/S and compiler
versions.  And each combination must be tested.

All of which means that a new release of a compiler might sit in
the box for months before we can use it.

So while it's true that there are compilers out there right now
that are close to ISO, that doesn't help some of us.

> The remaining different features still vary a bit. They tend to be
> something like partial specialization, "export", or Koenig lookup.
> However, there clearly is rapid convergence in the latest releases.

I look forward to writing product code that uses those things.
But it's not going to be any time soon.

-- David R. Tribble, dtribble@technologist.com --
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 8 Mar 1999 18:52:08 GMT, Al Stevens <alstevens@midifitz.com> wrote:
>Steve Clamage wrote in message <7bugfd$nin$1@engnews1.eng.sun.com>...
>
>>The C++ language standard is not a spec designed for manufacture,
>>nor should it be. It provides a definition of the language, not
>>a specification for a compiler.
>
>One could draw quite the opposite conclusion by reading the first sentence
>of 1.1 Scope.

It says:

"This International Standard specifies requirements for
implementations of the C++ programming language."

That's what Steve said: The standard provides a definition of the
language, which a conforming compiler must observe. The standard does
not specify how to write a conforming compiler. (That would be a whole
other kettle of wax, and ball of fish too.)

Herb


---
Herb Sutter (mailto:hsutter@peerdirect.com)

PeerDirect Inc.     2695 North Sheridan Way, Suite 150
www.peerdirect.com  Mississauga Ontario Canada L5K 2N6
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Steve Clamage <Stephen.Clamage@Sun.COM> wrote:
: Oleg Zabluda <zabluda@math.psu.edu> writes:
: >Steve Clamage <Stephen.Clamage@sun.com> wrote:

: >: It seems to me the best chance for a reference implementation
: >: along the lines I have suggested would have been g++ or egcs.
: >: The much-vaunted "bazaar" development model should have
: >: produced a reference implementation. A number of contributors
: >: to g++/egcs are long-time C++ committee members. So why didn't
: >: it happen? Does anyone from that bazaar want to comment?
: >: (I think the answer is the difficulty of producing a C++
: >: implementation -- the missing features are still being worked on.)

: >g++ has never followed the bazaar development model. g++ development
: >was stalled for about 2 years partially because of this. That's why
: >egcs project came to life. egcs is about 1.5 years old, yet it
: >has done more in this time then SunPro did in the last 5 years with
: >the much-vaunted non-bazaar development model.

: I didn't mean my comments as critical. I apologize if they
: sounded that way.

: I was asking an honest question. A reference implementation of
: C++ would require an investment by a commercial enterprise
: that thought it profitable or cost-effective (unlikely), or a
: volunteer effort.  The egcs project represents such a pool of
: volunteers, with development underway.

Egcs project as a whole does represent a significant pool of
volunteers. However, for better or worse, there are very little
unpaid volunteers doing C++ front-end work for egcs. Almost all
the work is done by commercial companies. Mostly Cygnus and Mark
Mitchell Consulting. There are more volunteers working on the
standard C++ library, however most of the work is being done by
commercial companies and paid contractors as well.

However, I am pretty sure that the involved commercial entities
will produce a reference implementation in time, just like they
did with gcc before, simply because they are constrained only by
the available resources, and their own technical judgement but
not (yet) by wierd PSB marketing decisions.

Oleg.
--
Life is a sexually transmitted, 100% lethal disease.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James.Kanze@dresdner-bank.com wrote:
: In article <7c342g$nsf@marianna.psu.edu>,
:   zabluda@math.psu.edu (Oleg Zabluda) wrote:

: > egcs is about 1.5 years old, yet it
: > has done more in this time then SunPro did in the last 5 years with
: > the much-vaunted non-bazaar development model.

: Do you have any actual tests to back this up.  I've not had the occasion
: to do any actual evaluations of egcs, so I can't prove you wrong.  But
: I've heard such claims so often in the past, and every time I did a real
: evaluation (test suites, etc.), Sun came out the best on the Sparc
: platforms.

Apparantly the answer will depend on what I mean by "more" and
you mean by "best", as you've desribed so many times before. The
situation before 5.0 (as of the end of 1998) was very clear-cut
for me. Egcs supports all of the standard C++ except the 'export',
while SunPro 4.2 can only be described as archaic at best.

It doesn't support 'bool', string literals have type 'char*'
(very funny interaction with CORBA, as I am sure you know), only
embryonic template support, no namespace support, totally messed
up handling of the top-level qualifiers, partial disregard to
scoping (again very funny interaction with CORBA when ::string can
be parsed as CORBA::string) [1], RogueWave stuff is dropped into
the system(!) include directory, and tons and tons of bugs on
top of it all.

Since the release of 5.0 the situation is temporarily is not so
clear cut anymore, but will be soon enough, just like it was
in the past.

[1] Check this out:

#include "CORBA.h"
::string str = "aaaa";

(core dump)

: To date, I've seen no solid empirical evidence to support the bazaar
: model,

Lesee... STL as we currently know it was developped with
the bazaar model.

: and it certainly violates every known rule of software process.

Fred Brooks' wisdom has it's limitations. For example, although it
does apply to groups of 100-1000 people working for the same employer,
it simply does not apply to the groups of millions of people hooked
up to the Internet.

Most importantly though, the bazaar model is making an excellent use
of (although it doesn't have a monopoly to) a well forgotten rule of
software process: Stand on each other's shoulders, not on each other's
toes.

Oleg.
--
Life is a sexually transmitted, 100% lethal disease.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36E5BB00.F70A0AEC@wizard.net>,
  James Kuyper <kuyper@wizard.net> wrote:
>
> Martin von Loewis wrote:
> >
> > James.Kanze@dresdner-bank.com writes:
> >
> > > I have been told (indirectly) that there are at least two which are
> > > shipping: Visual Age C++ 4.0 and Sun CC 5.0.  I've not been able to
> > > verify personally whether either are "100% compliant"; maybe someone
> > > from IBM or Sun would care to confirm.
> >
> > I work for neither information, but here references to some claims of
> > these organizations:
> >
> > Visual Age: Full support of the ANSI/ISO C++ Standard, including class
> >             library support
> > (http://www.software.ibm.com/ad/visualage_c++/)
>
> There's no weasel-wording about their claims of standard compliance, but
> also very little detail. That tends to make me suspicious.

The exact quote is "VisualAge C++ supports the C++ language and
libraries specified in the ANSI/ISO standard adopted in July 1998."
That's *not* the final standard.  On the other hand, it is very close;
I'm pretty sure it contained "export", for example.

> > Visual Workshop C++: Support for ANSI/ISO C++
> >                      (ISO/IEC14882-1998 C++ language standard)
> > (http://www.sun.com/workshop/visual/)
>
> That site merely says that it "Supports" the standard. It gives a lot
> more detail, if you look into the white papers, and some of that detail
> indicates incomplete compliance:
>
> "The C++ standard has some new rules for templates that make old code
> nonconforming, particularly code involving the use of the new keyword
> typename. The 5.0 compiler does not yet enforce these rules, but does
> recognize this keyword."

If the compiler supports all code conformant to the standard, and is
just not enforcing new rules, no problem.  If it is just recognizing the
new keyword, but doing nothing about it, however, "supports" is not in
order.

> and
>
> "When the export keyword and exported templates are implemented in a
> future compiler version, ..."

In sum, they don't support the entire standard.  Export is, IMHO, a
critical part of the standard; it gives us back (at least partially)
separate compilation of templates, which the committee more or less took
away.  (On the other hand, Sun CC has always supported separate
compilation of templates, even without export.  So perhaps they just
didn't think it necessary.  "#define export", and compile away.  Earlier
versions of Visual Age also supported separate compilation of templates,
so perhaps they simply allow the keyword, and ignore it:-).)

> and
>
> "The C++ standard allows using both the <NAME.h> and <cNAME> versions of
> the standard C headers in the same compilation unit. Although you
> probably would not do this on purpose, it can happen when you include,
> for example, <cstdlib> in your own code, and some project header you use
> includes <stdlib.h>. On Solaris environments 2.5.1 and 2.6, this mixing
> does not work for some headers, particularly for the
> <wchar.h>/<cwchar> and <wctype.h>/<cwctype> header pairs."

And on Solaris environments 2.7?

> Still, if that's the entirety of the non-compliance, it's not too bad!
> The prices are, however, way out of my ballpark.

N'est-ce pas?  Sun used to sell the compilers, unbundled, for a
reasonable price.  Now, however, they only sell them bundled with a IDE
and a debugger, at the top price, and I have problems getting an upgrade
on my maintanance contract because the product I bought doesn't exist
any more.

In the end, of course, the vendors are following the market.  If Sun
invests more to develop an IDE or on optimizing, than on implementing
export, it's because their customers don't care as much about export as
they do about IDE's or optimized code.  While I find export essential, I
seem to be an exception (and with the purchase of about one compiler
every three/four years, I'm hardly what you could call a major
customer).

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient   e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh   ttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 17946

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James.Kanze@dresdner-bank.com writes:

>In article <7c1ss5$35p$1@engnews1.eng.sun.com>,
>  Stephen.Clamage@sun.com (Steve Clamage) wrote:

>> With some notable exceptions, most recent compilers do a good job
>> of supporting the C++ standard. The exceptions are those vendors
>> who have serious compatibility issues that compliance would break,
>> or who have other issues that are more important to them than
>> standards compliance.

>But that means every serious vendor:-).  No serious vendor will ignore
>backwards compatibility, and break the clients code.  And surely quality
>is more important that compliance -- to a certain degree, so is a
>minimum level of performance.

It's a tough problem for C++ implementers, and for users with a large
code base, no question. At Sun, we've attempted to address both the
compatibility and standards-conformance issues simultaneously.  Our
intent is to provide a migration path from ARM-style to standard code.

Sun C++ 5.0 provides a compiler flag to select compatibility mode
(compiles the same code as our ARM-compliant compiler) or standard-
conforming mode. We supply two sets of headers and libraries:
compatibility and standard mode. We provide a Migration Guide that
explains how older code needs to be modified to work with a standard-
conforming compiler. We also provide a version of "classic" iostreams
compiled for use in "standard" mode.

I hope Sun is not alone in trying to address these conflicting
requirements.

--
Steve Clamage, stephen.clamage@sun.com
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James.Kanze@dresdner-bank.com writes:

>The exact quote is "VisualAge C++ supports the C++ language and
>libraries specified in the ANSI/ISO standard adopted in July 1998."
>That's *not* the final standard.

Yes it is. The date might be confusing. Here's the approximate
chronology:

November 1997:
    C++ Committee approves Final Draft International Standard.
December 1997 or January 1998:
    FDIS goes to ISO for approval by member nations.
April 1998:
    ISO member nation voting on FDIS begins.
June 1998:
    Voting closes.
July 1998:
    Voting results announced: FDIS passes, becomes the International
    Standard.
September 1998:
    Standard officially published.

If the process seems glacially slow to you, imagine how it seemed
to us on the committee who were wondering whether we were done!
The wonderful book title "Waiting to Exhale" comes to mind.

The only differences between the version approved in Nov 1997
and the version that was voted on by ISO member nations is the
correction of some minor typos and formatting glitches. I believe
the published standard is identical to the one that was first sent
to ISO. If not identical, the differences are not substantive.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James.Kanze@dresdner-bank.com writes:

> > > Visual Age: Full support of the ANSI/ISO C++ Standard, including class
> > >             library support
> > > (http://www.software.ibm.com/ad/visualage_c++/)
[...]

> The exact quote is "VisualAge C++ supports the C++ language and
> libraries specified in the ANSI/ISO standard adopted in July 1998."
> That's *not* the final standard.  On the other hand, it is very close;
> I'm pretty sure it contained "export", for example.

Just for the sake of correctness: I did find the above quote exactly
at the referenced location. Where is your quote from (I really like to
know)?

Also, in what way is the final standard different from the *adopted*
one? I don't think ISO publishing can change the text that was voted
on just like that...

Regards,
Martin


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c6dbl$3u7$1@nnrp1.dejanews.com>, James.Kanze@dresdner-
bank.com writes
>The exact quote is "VisualAge C++ supports the C++ language and
>libraries specified in the ANSI/ISO standard adopted in July 1998."
>That's *not* the final standard.  On the other hand, it is very close;
>I'm pretty sure it contained "export", for example.

Now I am confused.  Why is that not the standard.  We shipped the
standard in Autumn 1997 and the Draft Standard was being voted on during
the spring of 98.  As it passed only typos could be changed after that
date.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Steve Clamage wrote:

> I was asking an honest question. A reference implementation of
> C++ would require an investment by a commercial enterprise
> that thought it profitable or cost-effective (unlikely), or a
> volunteer effort.  The egcs project represents such a pool of
> volunteers, with development underway.
>

Unfortuantely, standards compliance doesn't seem to be one
of their goals.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

The quotes at issue:

"This International Standard specifies requirements for
implementations of the C++ programming language"


and

"The standard does not specify how to write a conforming compiler,"


seem to me to be at odds with one another. What is a "conforming compiler"
if not one of those "implementations of the C++ programming language?" Of
course you can interpret either quote any way you wish.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Ron Natalie wrote in message <36E70BFD.74DD0CC8@sensor.com>...
>
> [ re EGCS ]
>
>Unfortuantely, standards compliance doesn't seem to be one
>of their goals.

What gives you that idea? EGCS is one of the most compliant compilers
available, and the EGCS and Libstdc++ teams are making a very credible
effort to bring it fully up to spec. What makes you think otherwise?

--
Ross Smith ................................... mailto:ross.s@ihug.co.nz
.............. The Internet Group, Auckland, New Zealand ..............
         "The award for the Most Effective Promotion of Linux
         goes to Microsoft."             -- Nicholas Petreley




[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

"Al Stevens" <alstevens@midifitz.com> writes:
>
> "This International Standard specifies requirements for
> implementations of the C++ programming language"
>
> and
>
> "The standard does not specify how to write a conforming compiler,"
>
> seem to me to be at odds with one another. What is a "conforming
> compiler" if not one of those "implementations of the C++
> programming language?" Of course you can interpret either quote any
> way you wish.

The two quotes are not at all at odds with one another. The first
quote talks about "requirements", the second about implementation,
i.e. the word "how" is the emphasis in the second quote, while the
first talks about the "what".

The point being made here is of great importance for software
engineering in the large, namely the distinction between requiremnts
and their realisations, between interfaces and their implementations,
between abstractions and their concrete representations.

A good standard unambiguously pins down the requirements, abstractions,
and interfaces, without unneccessarily restricting implementation
choices.

Joachim

--
joachim@kraut.bc.ca      (http://www.kraut.bc.ca)
joachim@mercury.bc.ca    (http://www.mercury.bc.ca)


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Al Stevens wrote:
>
> The quotes at issue:
>
> "This International Standard specifies requirements for
> implementations of the C++ programming language"
>
> and
>
> "The standard does not specify how to write a conforming compiler,"
>
> seem to me to be at odds with one another. What is a "conforming compiler"
> if not one of those "implementations of the C++ programming language?" Of
> course you can interpret either quote any way you wish.

The distinction you're missing is not between a "compiler" and an
"implementation"; a compiler can be an implementation. It is the
distiction between "requirements" and "how to write", which are very
different. I could have a requirement to sort a data set; that
requirement doesn't tell me how to write the sorting routine.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In this next-to-last year of the millennium, AllanW@my-dejanews.com
(AllanW@my-dejanews.com) wrote in article
<7c1tcg$6du$1@nnrp1.dejanews.com> in comp.std.c++:
>> I think none of us has ever seen a bug-free program of any size,
>
>Of course we have. Write a trivial program, such as one that reports
>the sum of all numbers on the command line.

Sorry -- ambiguities in English. I meant "program of any size" to mean
"sizeable program", i.e. one beyond the toy stage. You took the other
meaning, "of any size" = "no matter what the size". Of course in that
sense we have all seen bug-free programs. (That should have been clear
from what followed in my article, but I guess it wasn't.)

>> know by now that correctness is a sort of mathematical limit we can
>> approach as closely as we like, given sufficient time and money, but
>> never quite reach.
>
>Tell that to NASA. Reaching it is expensive, but possible.

You maintain that NASA software is bug free? On what do you base that
rather startling claim?

--
Stan Brown, Oak Road Systems, Cleveland, Ohio, USA
                                    http://www.mindspring.com/~brahms/
My reply address is correct as is. The courtesy of providing a correct
reply address is more important to me than time spent deleting spam.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <T8ZRTpADA$42EwRE@robinton.demon.co.uk>,
francis@robinton.demon.co.uk says...
> In article <MPG.114ccce424aa5d2d989b7c@netnews.worldnet.att.net>, Tony
> <Tony@ask.me> writes
> >Yeap, I was only hypothesizing.  Review is important.  I was just
> >wondering how smoothly the process ran and whether PM was being practiced
> >or if it was more like a bunch of techies arguing.
>
> In the early days the latter might have been a fair characterisation but
> you do not think we would have delivered if it had stayed that way, do
> you?

Dunno.  Seemed like it took a long time.

> It was really hard for some to grit their teeth and accept that
> time was a real constraint.

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c1pji$36t$1@nnrp1.dejanews.com>, AllanW@my-dejanews.com
says...
>
> In article <MPG.114a5752cd41bf19989b66@netnews.worldnet.att.net>,
>   Tony@ask.me (Tony) wrote:
> > In article <36DF3EB5.3CE14A6D@wizard.net>, kuyper@wizard.net says...
> > > From the anti-trust point of view, standardization is comparable to
> > > price-fixing. Price-fixing inhibits price-based competition;
> > > standardization inhibits feature-based competition.
> >
> > Standards should probably be well accepted practices and lowest-common-
> > denominator-like things where there is no question as to their motives.
> > From that foundational base, there is plenty of room for value-added
> > features and innovation.  What is sad, as you allude to, is the situation
> > where "standards" (de facto perhaps) eliminate choice of the user and
> > oppress potential developers (e.g. the GUI situation).
>
> I think that what Tony is getting at, is that intent matters. If
> ANSI/ISO didn't have exemptions in federal law, then someone could
> conceivably sue for anti-trust. However, such a lawsuit would
> surely fail if ISO could demonstrate that
>
>   1) The organization allowed anyone to become a member for a
>      relatively small investment
>   2) The organization had no ability nor desire to force anyone
>      to follow the standards
>   3) Even those companies that do follow the standards, are not
>      obligated to leave out additional features

OK, let me ask the tough question then: "Is the std complex _by_design_
(even if _indirectly_ manipulated to be so)?"

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Stan Brown wrote:
>
> In this next-to-last year of the millennium, AllanW@my-dejanews.com
> (AllanW@my-dejanews.com) wrote in article
> <7c1tcg$6du$1@nnrp1.dejanews.com> in comp.std.c++:
....
> >> know by now that correctness is a sort of mathematical limit we can
> >> approach as closely as we like, given sufficient time and money, but
> >> never quite reach.
> >
> >Tell that to NASA. Reaching it is expensive, but possible.
>
> You maintain that NASA software is bug free? On what do you base that
> rather startling claim?

I write NASA software. I can certify that it isn't. Just yesterday I
found a debugging printf() in the latest release of a utility library.
All delivered code that will be using that library is prohibited (by the
same people who designed that library!) from writing to stdout.

Of course, my software analyzes data after it comes down from the
satellite. The truly critical software is the flight operations software
that controls the satellite. Our launch has been delayed by nearly a
year because the original contractor for the flight ops software bungled
the job.

NASA does have the concept of man-rated software. That's software that
is considered to be reliable enough to justify risking a man's life on
it's reliability. I'm glad that I'm not required to produce man-rated
code. I've heard that it's considered a very serious fault if man-rated
software produces even a single compiler warning, the very FIRST time it
gets compiled.

Despite this, even man-rated software fails sometimes.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James Kuyper <kuyper@wizard.net> writes:

>The original idea on this thread was to produce a standard outside of
>ISO's normal slow, complicated procedures. The problem with that is that
>whatever power ISO does have, is the result of earning a reputation as
>an organization that produces well-considered standards with
>international participation in the process. I'm not saying that every
>ISO procedure is well-justified; I don't even know enough about the
>detailed procedures to comment on them. I am saying that slow, carefully
>considered consensus building is a good idea in this context, and that
>this proposal seems motivated by a desire to avoid precisely those
>features of the ISO process.

No, you've misunderstood.  By all means let the "free standards"
organization adopt a similarly carefully considered, consensus
building approach, and if that means being slow, so be it.
The aim is not to provide faster development, the aim is to ensure
that the final result of the process is freely available to all.

--
Fergus Henderson <fjh@cs.mu.oz.au>  |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>  |  of excellence is a lethal habit"
PGP: finger fjh@128.250.37.3        |     -- the last words of T. S. Garp.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 09 Mar 99 01:53:55 GMT, ark@research.att.com (Andrew Koenig) wrote:

} In article <7c07p7$bfu$2@remarQ.com>,
} Reality is a point of view <gjohnson@dream.season.com> wrote:
}
} > Any estimates on how many more months will pass before a
} > reasonably complete ISO C++ implementation appears?
}
} There are already two implementations on the market that claim
} to implement the entire standard.

Please name them.

Have fun,
Alex.

--

My email address is intentionally mangled to foil spambots.
Please remove the "---filter---" from the address for replying.
Sorry for the inconvenience.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

AllanW@my-dejanews.com wrote in message <7c1tcg$6du$1@nnrp1.dejanews.com>...
>In article <MPG.114b5f4319abb5b98973c@news.mindspring.com>,
>  brahms@mindspring.com (Stan Brown) wrote:
>>
>> Look at how many people try to discover whether something is legal C++ by
>> trying it in different compilers. It's easy to do and gives a clear
>> answer. The problem, as we know, is that though the answer is clear it
>> may not be correct. The reference implementation's answer would be clear
>> and correct (modulo any defect reports, which would also be clear and
>> unambiguous).
>
>Good point, but this also gives the reference implementation an extra
>requirement that commercial versions would not have: *ALL* invalid
>programs *MUST* fail to compile or execute correctly. i.e. we had better
>make sure that a program which casts a void* to a long and back again
>doesn't run without any diagnostic errors. Otherwise, we will have at
>least some programmers using the reference implementation to prove that
>this is legal, and then demand that the commercial versions do so as well.

I suspect this requirement would make such a reference implementation
impossible. Aren't there some kinds of undefined behaviour for which no
diagnostic is required because it's impossible to detect them in some
cases?

--
Ross Smith ................................... mailto:ross.s@ihug.co.nz
.............. The Internet Group, Auckland, New Zealand ..............
         "The award for the Most Effective Promotion of Linux
         goes to Microsoft."             -- Nicholas Petreley




[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

AllanW@my-dejanews.com writes:

>In article <7c0s08$576$1@engnews1.eng.sun.com>,
>  Stephen.Clamage@sun.com (Steve Clamage) wrote:
>> The main hurdle remains: who will do the work? It's easy for
>> non-participants to say that "someone" should do it. But
>> non-participants have already shown that they themselves
>> are unwilling or unable to contribute.

>For some of us, the unable is much bigger than the unwilling.

>Let me plead "poor programmer" for a while. ...

I did not mean my comment to be generally derogatory, but
primarily a statement of fact.

Standards work is hard, exacting, and detailed. The work is done
outside your regular job, unless you are lucky enough to have
an employer who makes it your job.  Few people have both
the temperament and the resources available to work consistently
on standards. Many fine, smart, capable, effective people
don't fall into that category. (Luckily for us, some do.)

But we sometimes see comments from individuals who do not
participate in the standards process and who know little or
nothing about it, taking the committee to task for imagined
omissions or malfeasance, or insisting the committee take on
significant extra work because they think it is a good idea.

I find such comments objectionable, and that is the sort of
person I had in mind in the paragraph quoted above.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36e4e372.87956354@news.netvision.net.il>,
  zivc@peach-networks.com (Ziv Caspi) wrote:
> On 09 Mar 99 01:57:20 GMT, AllanW@my-dejanews.com wrote:
> [...]
> >A "reference implementation" that does not generate code would amount
> >to little more than a "lint" program.
>
> Surely you jest.

No, this is really my opinion.

> Writing a conforming front end for a C++ compiler is
> a very large amount of work. In all cases I know of, the task of
> writing a front end (for languages much simpler than C++) was
> substantially larger than writing the back end, optimizers &c.

Agreed.

However difficult such a tool would be to generate, I still believe
that if it failed to generate executable code or equivalent, it
would not have value.

Suppose your next automobile was fully functional, except that the
wheels didn't turn. Surely making the engine, transmission, and
air conditioning function properly is much more difficult than
connecting the drive train. And yet, despite all this hard work,
the car is still useless.

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c07p7$bfu$2@remarQ.com>,
  gjohnson@dream.season.com (Reality is a point of view) wrote:
>  +---- bs@research.att.com wrote (7 Mar 1999 17:35:14 GMT):
>  | Now is the time to plan and experiment for the future, not to quibble
>  | over the past.
>  +----
>
> Any estimates on how many more months will pass before a
> reasonably complete ISO C++ implementation appears?

I have been told (indirectly) that there are at least two which are
shipping: Visual Age C++ 4.0 and Sun CC 5.0.  I've not been able to
verify personally whether either are "100% compliant"; maybe someone
from IBM or Sun would care to confirm.

Realistically, both have version numbers ending with 0, so some caution
should be exercised with regards to the stability.  I have found that in
the past, both of these companies do take quality seriously -- there
will be bugs, but I wouldn't expect too many show-stoppers.  On the
other hand, I don't think I'd plunge right in with an application which
stresses the compiler if the company depended on it.

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient   e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh   ttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 17946

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c1h9g$ki$2@remarQ.com>,
  gjohnson@dream.season.com (Reality is a point of view) wrote:

>  | I can do most of what I need to do using EDG-based compilers, GNU's EGCS,
>  | MS and Borland compilers, and Sun just released their 5.0 which is claimed
>  | to be pretty close.
>  |
>  | The remaining different features still vary a bit. They tend to be
something
>  | like partial specialization, "export", or Koenig lookup. However, there
>  | clearly is rapid convergence in the latest releases.
>  +----
>
> So there aren't any major pieces that are commonly left
> unimplemented, or partially implemented?  Are the variant bits
> confined to little used areas?

Well, IMHO, export would qualify as a major feature, necessary for any
serious use of templates.  If a compiler didn't support it, then I would
certainly not qualify it as "conforming".

And obviously, if a major feature is not widely supported, it will be
little used.

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient   e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh   ttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 17946

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bvk07$qj4@bgtnsc02.worldnet.att.net>,
  "Andrei Alexandrescu" <andrewalex@hotmail.com> wrote:
> Stan Brown wrote in message ...
> [snip]
>
> Reading all this I cannot stop myself thinking of John Lilley's pccts-based
> C++ parser.
>
> As you may already know, pccts (Purdue Compiler Construction Tool Set) is a
> lex+yacc like tool for building parsers. (http://www.antlr.com)

Could this name be wrong? http://www.antlr.com -- could not connect to
server.

 [Moderator's note: the correct URL is <http://www.antlr.org>.
        -moderator (fjh).]                               ^^^

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c1ss5$35p$1@engnews1.eng.sun.com>,
  Stephen.Clamage@sun.com (Steve Clamage) wrote:

> With some notable exceptions, most recent compilers do a good job
> of supporting the C++ standard. The exceptions are those vendors
> who have serious compatibility issues that compliance would break,
> or who have other issues that are more important to them than
> standards compliance.

But that means every serious vendor:-).  No serious vendor will ignore
backwards compatibility, and break the clients code.  And surely quality
is more important that compliance -- to a certain degree, so is a
minimum level of performance.

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient=E9e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh=FCttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 179=
46

-----------=3D=3D Posted via Deja News, The Discussion Network =3D=3D----=
------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own  =
 =20
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James Kuyper (kuyper@wizard.net) wrote:
: The original idea on this thread was to produce a standard outside of
: ISO's normal slow, complicated procedures.

Perhaps that was what some people were getting at. But others of us are
more interested in getting rid of restrictions on useful things that can be
done with the standard now that its been approved. Getting ISO out of the
process (or mainline thereof) is one way to do that, but perhaps not the
best way.

-Z-


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <4+t7TkASot42Ew1R@robinton.demon.co.uk>,
  Francis Glassborow <francisG@robinton.demon.co.uk> wrote:
> In article <36e1c862.2367323@nr1.toronto.istar.net>, Herb Sutter
> <hsutter@peerdirect.com> writes
> >Come and find out! :) We're meeting in Dublin next month. See this
> >newsgroup's FAQ to learn how you can get involved.
>
> Getting involved in standardisation at both national (UK) and
> international level is the single best decision I ever made with respect
> to improving my programming skills.

I can second this.  The biggest win of my (limited) participation was
the people I met, but the second was what I learned.

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient   e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh   ttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 17946

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c0s08$576$1@engnews1.eng.sun.com>,
  Stephen.Clamage@sun.com (Steve Clamage) wrote:

> It seems to me the best chance for a reference implementation
> along the lines I have suggested would have been g++ or egcs.
> The much-vaunted "bazaar" development model should have
> produced a reference implementation. A number of contributors
> to g++/egcs are long-time C++ committee members. So why didn't
> it happen? Does anyone from that bazaar want to comment?
> (I think the answer is the difficulty of producing a C++
> implementation -- the missing features are still being worked on.)

Although I think that much can be said about the bazaar model, in this
case, it is not at fault.  In the end, the people working on egcs have
much the same priorities as the compiler writers at Sun and Microsoft:
they have a community of users, and they are responsive the the needs of
that community.  If egcs doesn't support the export keyword at present,
at least part of the reason is that the egcs users don't consider it
important.  Or at least, not as important as other things: bazaar model
or no, the resources available to develop egcs are not infinite.

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient   e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh   ttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 17946

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Oleg Zabluda <zabluda@math.psu.edu> writes:

>Steve Clamage <Stephen.Clamage@sun.com> wrote:

>: It seems to me the best chance for a reference implementation
>: along the lines I have suggested would have been g++ or egcs.
>: The much-vaunted "bazaar" development model should have
>: produced a reference implementation. A number of contributors
>: to g++/egcs are long-time C++ committee members. So why didn't
>: it happen? Does anyone from that bazaar want to comment?
>: (I think the answer is the difficulty of producing a C++
>: implementation -- the missing features are still being worked on.)

>g++ has never followed the bazaar development model. g++ development
>was stalled for about 2 years partially because of this. That's why
>egcs project came to life. egcs is about 1.5 years old, yet it
>has done more in this time then SunPro did in the last 5 years with
>the much-vaunted non-bazaar development model.

I didn't mean my comments as critical. I apologize if they
sounded that way.

I was asking an honest question. A reference implementation of
C++ would require an investment by a commercial enterprise
that thought it profitable or cost-effective (unlikely), or a
volunteer effort.  The egcs project represents such a pool of
volunteers, with development underway.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James.Kanze@dresdner-bank.com writes:

> I have been told (indirectly) that there are at least two which are
> shipping: Visual Age C++ 4.0 and Sun CC 5.0.  I've not been able to
> verify personally whether either are "100% compliant"; maybe someone
> from IBM or Sun would care to confirm.

I work for neither information, but here references to some claims of
these organizations:

Visual Age: Full support of the ANSI/ISO C++ Standard, including class
            library support
(http://www.software.ibm.com/ad/visualage_c++/)

Visual Workshop C++: Support for ANSI/ISO C++
                     (ISO/IEC14882-1998 C++ language standard)
(http://www.sun.com/workshop/visual/)

Regards,
Martin


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c342g$nsf@marianna.psu.edu>,
  zabluda@math.psu.edu (Oleg Zabluda) wrote:
> Steve Clamage <Stephen.Clamage@sun.com> wrote:
>
> : It seems to me the best chance for a reference implementation
> : along the lines I have suggested would have been g++ or egcs.
> : The much-vaunted "bazaar" development model should have
> : produced a reference implementation. A number of contributors
> : to g++/egcs are long-time C++ committee members. So why didn't
> : it happen? Does anyone from that bazaar want to comment?
> : (I think the answer is the difficulty of producing a C++
> : implementation -- the missing features are still being worked on.)
>
> g++ has never followed the bazaar development model. g++ development
> was stalled for about 2 years partially because of this. That's why
> egcs project came to life. egcs is about 1.5 years old, yet it
> has done more in this time then SunPro did in the last 5 years with
> the much-vaunted non-bazaar development model.

Do you have any actual tests to back this up.  I've not had the occasion
to do any actual evaluations of egcs, so I can't prove you wrong.  But
I've heard such claims so often in the past, and every time I did a real
evaluation (test suites, etc.), Sun came out the best on the Sparc
platforms.

To date, I've seen no solid empirical evidence to support the bazaar
model, and it certainly violates every known rule of software process.

--
James Kanze                         mailto: James.Kanze@dresdner-bank.com
Conseils en informatique orient   e objet/
                              Beratung in industrieller Datenverarbeitung
Ziegelh   ttenweg 17a, 60598 Frankfurt, Germany    Tel. +49 (069) 263 17946

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Martin von Loewis wrote:
>
> James.Kanze@dresdner-bank.com writes:
>
> > I have been told (indirectly) that there are at least two which are
> > shipping: Visual Age C++ 4.0 and Sun CC 5.0.  I've not been able to
> > verify personally whether either are "100% compliant"; maybe someone
> > from IBM or Sun would care to confirm.
>
> I work for neither information, but here references to some claims of
> these organizations:
>
> Visual Age: Full support of the ANSI/ISO C++ Standard, including class
>             library support
> (http://www.software.ibm.com/ad/visualage_c++/)

There's no weasel-wording about their claims of standard compliance, but
also very little detail. That tends to make me suspicious.

> Visual Workshop C++: Support for ANSI/ISO C++
>                      (ISO/IEC14882-1998 C++ language standard)
> (http://www.sun.com/workshop/visual/)

That site merely says that it "Supports" the standard. It gives a lot
more detail, if you look into the white papers, and some of that detail
indicates incomplete compliance:

"The C++ standard has some new rules for templates that make old code
nonconforming, particularly code involving the use of the new keyword
typename. The 5.0 compiler does not yet enforce these rules, but does
recognize this keyword."

and

"When the export keyword and exported templates are implemented in a
future compiler version, ..."

and

"The C++ standard allows using both the <NAME.h> and <cNAME> versions of
the standard C headers in the same compilation unit. Although you
probably would not do this on purpose, it can happen when you include,
for example, <cstdlib> in your own code, and some project header you use
includes <stdlib.h>. On Solaris environments 2.5.1 and 2.6, this mixing
does not work for some headers, particularly for the
<wchar.h>/<cwchar> and <wctype.h>/<cwctype> header pairs."

Still, if that's the entirety of the non-compliance, it's not too bad!
The prices are, however, way out of my ballpark.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes

 > > The ISO standard was already delayed about two years (while the
 > > STL was being rolled into it); would anyone have wanted to wait even
 > > longer for final acceptance?
 >
 > So why create a mononlithic standard anyway?  (Analogy: Mach and BSD UNIX
 > kernels).

I do not believe that the STL delayed the standard for two years.
There were several issues that stopped us from declaring the standard
"done." The integration of the STL was only one of those. Maybe we would
have completed a few month earlier without the STL, but I do not believe
we would have saved many months.

Shipping a standard without a significant standard library would have
been to repeat my original mistake with C++: Everybody would have to

 (1) write based on the C library only, or
 (2) rely on one of the many commercial foundation libraries, or
 (3) start trying to design and implement string, vector, list ,etc.

This would have been to perpetuate chaos.

What the C++ community needs is more standard libraries, not fewer.

IMO, the committee showed commendable restraint and specified only
the few standard library facilities that they could handle. We can argue
whether they specified one too many (say, valarray) or one too few (say,
hash_map), but I think that (given the available time, resources, and
proposals) the committee produced something of the right magnitude.

Now is the time to plan and experiment for the future, not to quibble
over the past.

 - Bjarne
Bjarne Stroustrup - http://www.research.att.com/~bs




[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7brpoh$mlc$1@engnews1.eng.sun.com>,
Stephen.Clamage@Eng.Sun.COM says...
> Tony@ask.me (Tony) writes:
>
> >In article <eOMyPIAoF932EwrY@robinton.demon.co.uk>,
> >francis@robinton.demon.co.uk says...
> >> In article <MPG.1148da60366d462f989b53@netnews.worldnet.att.net>, Tony
> >> <Tony@ask.me> writes
> >> >That should be easily doable just by using the money that is
> >> >currently being used to congregate (someone said 8k per person per
> >> >trip!).  Have all interactions occur on the net, dropping expenses to
> >> >almost nil.  :)
> >>
> >> Sorry (yes I see the smiley) but that really does not work people still
> >> need to meet (and much of the graft is already done over the Internet).
> >> Anyway how does that cut the cost of time.  The real cost is not the
> >> airfares and hotel bills but in the work time spent away from the
> >> employers product.
>
> >My guess is that like any committee project, too much is being discussed
> >in too much detail.
>
> If you don't discuss everything in great detail, you don't produce
> a standard of acceptable quality. This is a language standard.
> Every sentence matters. Every language detail must be covered.

My point was that _everybody_ shouldn't be doing _everything_.  Sounds
like higher level of abstraction for responsibilities may be in order if
everything is proceeding at level 1 (highest level of detail).

> >Is the project management of these events any good?
>
> This is an all-volunteer effort. If you are volunteering to take
> over the project and produce a better result faster, I will be
> happy, nay eager, to resign as chair in your favor.

Well thanks for the pseudo-vote-O-confidence, but I'm not in a position
to volunteer for anything right now.  The "chair" is equivalent to
"project manager" then you say?  Most times I have observed that what is
being called the PM role is actually being filled by a Project Leader,
hence the PM tasks are not being done.  Just throwing out some ideas
here.

> If you haven't been part of a standards process, you don't appreciate
> where many of the difficulties lie. It isn't anything like producing a
> commercial product.

What are they?  Why not?

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.114b5f4319abb5b98973c@news.mindspring.com>, Stan Brown
<brahms@mindspring.com> writes
>I think a reference implementation would be wonderful to have.
>

So do I, but I also know how many hours of work from several
exceptionally talented programmers went into developing a model compiler
for C.  During its development numerous bugs within the C standard were
located (each of those took many hours of work in ensuring that they
really were cases of the C standard being inconsistent).  Even after all
that work, new problems with the C standard surfaced.  Almost all, if
not actually all, were subtle and of interest only to language
theoreticians.  However the existence of such nits ensures that any
process of program proving must fail (the language definition is
inconsistent or incomplete).

As such a tool (model compiler) has considerable commercial value to
those who need the highest quality code (many more than actually think
they do) I feel fairly certain that someone is attempting to produce
such a tool.  However look at a five figure price per seat as a
reasonable price tag.



Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:

>In article <7brork$lki$1@engnews1.eng.sun.com>, Stephen.Clamage@sun.com
>says...
>>
>> Tony@ask.me (Tony) writes:
>>
>> >In theory, the two (specification and implementation) would
>> >_evolve_ concurrently and should be in complete agreement at the end of
>> >the project.  The influence of the implementation on the specification, I
>> >believe, would result in a standard of higher quality.
>>
>> Now we are back to assuming that you can create an implementation
>> known to be bug-free for something as complex as C++ in a reasonable
>> time frame. If that were possible, we'd have complete and correct
>> compilers now.
>>
>> I claim it isn't possible in any time frame of interest.

>Then to me that sounds like the spec is not adequately "designed for
>manufacture" and needs more work.  Just the kind of thing concurrent
>engineering should prevent.

The C++ language standard is not a spec designed for manufacture,
nor should it be. It provides a definition of the language, not
a specification for a compiler.

The language definition must define the form and meaning of a valid
program, as well as other requirements on implementations.

A program specification serves a rather different purpose.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36e1c862.2367323@nr1.toronto.istar.net>,
hsutter@peerdirect.com says...
>
> Tony:
> >>My guess is that like any committee project, too much is being discussed
> >>in too much detail.
>
> Instead of guessing, why not attend and find out? You would be
> welcome.

I have enough to do right now just building things with the language.
I'll trust you guys to keep up the good work.

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36e1c862.2367323@nr1.toronto.istar.net>, Herb Sutter
<hsutter@peerdirect.com> writes
>Come and find out! :) We're meeting in Dublin next month. See this
>newsgroup's FAQ to learn how you can get involved.

Getting involved in standardisation at both national (UK) and
international level is the single best decision I ever made with respect
to improving my programming skills.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Stan Brown wrote in message ...
[snip]

Reading all this I cannot stop myself thinking of John Lilley's pccts-based
C++ parser.

As you may already know, pccts (Purdue Compiler Construction Tool Set) is a
lex+yacc like tool for building parsers. (http://www.antlr.com)

John Lilley has taken the grammar of a draft standard and expressed it in
pccts' EBNF format, together with symbol table management. He's put
everything in the public domain. All in all, his tool can parse a C++ source
file and reveal errors (no code generated). I tested it and ot worked all
right.

The grammar has about 1000 lines of human-readable, well-written code (that
makes intensive use of STL and exceptions). It's an excellent piece of work.
However, some C++ constructs are not supported, but due to the clarity and
terseness of the grammar, it can be easily extended.

One things hinders this extraordinary acheivement: pccts is not developed
anymore. Its author, Terence Parr, hates C++ so he went for developing
antlr, a Java-based parser generator. antlr's grammar is not as powerful as
pccts' one. John Lilley told me in an email that porting his C++ parser to
antlr would be a major effort.
Besides, John himself is not interested in developing his parser anymore.

Andrei
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

 +---- bs@research.att.com wrote (7 Mar 1999 17:35:14 GMT):
 | Now is the time to plan and experiment for the future, not to quibble
 | over the past.
 +----

Any estimates on how many more months will pass before a
reasonably complete ISO C++ implementation appears?

--
Gary Johnson     gjohnson@season.com
Privacy on the net is still illegal.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bvk07$qj4@bgtnsc02.worldnet.att.net>, Andrei Alexandrescu
<andrewalex@hotmail.com> writes
>John Lilley has taken the grammar of a draft standard and expressed it in
>pccts' EBNF format, together with symbol table management. He's put
>everything in the public domain. All in all, his tool can parse a C++ source
>file and reveal errors (no code generated). I tested it and ot worked all
>right.

The problem is that there is quite a bit more in the way of constraints
etc. that is not covered by the grammar.  For example, the full order of
evaluation of an expression is not covered by the grammar.

Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.114b62ed93a69904989b6a@netnews.worldnet.att.net>,
Tony <Tony@ask.me> wrote:
> In article <F850xy.6po@research.att.com>, ark@research.att.com says...

> > > That should be easily doable just by using the money that is
> > > currently being used to congregate (someone said 8k per person per
> > > trip!).  Have all interactions occur on the net, dropping expenses to
> > > almost nil.  :)

> > The trouble with statements of the form ``that should be easily doable''
> > is that it doesn't mention who would do it, or why they would be
> > interested in doing so.

> You obviously didn't read the whole post which read like: "save in one
> area, apply to another".

Yes, I did read it.
--
    Andrew Koenig
    ark@research.att.com
    http://www.research.att.com/info/ark



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bugfd$nin$1@engnews1.eng.sun.com>, Stephen.Clamage@sun.com
says...
> >Then to me that sounds like the spec is not adequately "designed for
> >manufacture" and needs more work.  Just the kind of thing concurrent
> >engineering should prevent.
>
> The C++ language standard is not a spec designed for manufacture,
> nor should it be. It provides a definition of the language, not
> a specification for a compiler.

Still, to some degree the analogy holds.  Again, being blind to the
eventual use of the standard as an input to compiler production shouldn't
be ignored.

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <l16Mh5ANyW42EwG5@robinton.demon.co.uk>,
francis@robinton.demon.co.uk says...
>
> In article <MPG.114a583e67633f36989b67@netnews.worldnet.att.net>, Tony
> <Tony@ask.me> writes
> >My guess is that like any committee project, too much is being discussed
> >in too much detail.  Is the project management of these events any good?
>
> A requirement for a standard is that those involved get as close to
> consensus as possible.  For something as important as a standard that is
> actually important.  You probably have little idea as to how often
> serious errors have been prevented by the sharp eyes of a single person.

Yeap, I was only hypothesizing.  Review is important.  I was just
wondering how smoothly the process ran and whether PM was being practiced
or if it was more like a bunch of techies arguing.

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

brahms@mindspring.com (Stan Brown) writes:

>In this next-to-last year of the millennium, Stephen.Clamage@sun.com
>(Steve Clamage) wrote in article <7brork$lki$1@engnews1.eng.sun.com> in
>comp.std.c++:
>>Now we are back to assuming that you can create an implementation
>>known to be bug-free for something as complex as C++ in a reasonable
>>time frame. If that were possible, we'd have complete and correct
>>compilers now.

>Not necessarily (see below).

>>I claim it isn't possible in any time frame of interest.

>Literally true, Steve, but I think it's beside the point.

I think it is exactly the point in this specific case.

Tony@ask.me suggested that there should be a reference implementation
of C++. In the process of clarifying exactly what he meant, I
concluded that he had in mind an implementation process that
went on concurrently with the standard development, such that
every feature would have been tried out before becoming part
of the specification. I said I thought that was a good idea.

Tony said (I'm paraphrasing) that no, he meant a complete
implementation developed in parallel with the standard so that
the end of the process, we'd have a complete and correct
implementation along with a complete and correct standard.
It was that result I claimed was not possible.

In fact, various implementers did work on new features as they
were proposed. In the course of doing so, many specification flaws
were discovered and corrected.

Unfortunately, not all features were fully implemented and tested
before the standard was complete; a few features of the standard
still have not been fully implemented by anyone, AFAIK (template
export semantics and template template parameters in particular).

Some committee members argued that we shouldn't specify a feature
until we had some implementation experience, but they could not
sustain a majority vote in all cases.

The main hurdle remains: who will do the work? It's easy for
non-participants to say that "someone" should do it. But
non-participants have already shown that they themselves
are unwilling or unable to contribute. The participating
committee members do as much as they are able, particularly
the commercial implementers who have the largest stake.

But as Stan points out in good arguments that I snipped,
commercial implementers often have other priorities, compliance
with the standard in some cases not being near the top of
the list. Nevertheless, some 3rd party implementors have as
their main selling point compliance with the standard.
They have certainly done their best to implement as much as
possible, as quickly as possible.

It seems to me the best chance for a reference implementation
along the lines I have suggested would have been g++ or egcs.
The much-vaunted "bazaar" development model should have
produced a reference implementation. A number of contributors
to g++/egcs are long-time C++ committee members. So why didn't
it happen? Does anyone from that bazaar want to comment?
(I think the answer is the difficulty of producing a C++
implementation -- the missing features are still being worked on.)

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:


>In article <7bugfd$nin$1@engnews1.eng.sun.com>, Stephen.Clamage@sun.com
>says...
>> >Then to me that sounds like the spec is not adequately "designed for
>> >manufacture" and needs more work.  Just the kind of thing concurrent
>> >engineering should prevent.
>>
>> The C++ language standard is not a spec designed for manufacture,
>> nor should it be. It provides a definition of the language, not
>> a specification for a compiler.

>Still, to some degree the analogy holds.  Again, being blind to the
>eventual use of the standard as an input to compiler production shouldn't
>be ignored.

Of course it wasn't ignored. Most implementers of C++, commercial
and otherwise, were represented on the C++ committee.

I recommend you become informed about the actual requirements for
producing an ISO language standard, and the procedures (however
flawed they may be) actually used by the C++ committee.

You could then provide some constructive criticism instead of
flinging random accusations in all directions.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Andrei Alexandrescu wrote:
>
> John Lilley has taken the grammar of a draft standard and expressed it in
> pccts' EBNF format, together with symbol table management. He's put
> everything in the public domain.

Can you give the location where this C++ grammar is available?

--
Best regards,
Biju Thomas
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.114ccce424aa5d2d989b7c@netnews.worldnet.att.net>, Tony
<Tony@ask.me> writes
>Yeap, I was only hypothesizing.  Review is important.  I was just
>wondering how smoothly the process ran and whether PM was being practiced
>or if it was more like a bunch of techies arguing.

In the early days the latter might have been a fair characterisation but
you do not think we would have delivered if it had stayed that way, do
you?  It was really hard for some to grit their teeth and accept that
time was a real constraint.



Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

gjohnson@dream.season.com (Reality is a point of view) writes:

 >  +---- bs@research.att.com wrote (7 Mar 1999 17:35:14 GMT):
 >  | Now is the time to plan and experiment for the future, not to quibble
 >  | over the past.
 >  +----
 >
 > Any estimates on how many more months will pass before a
 > reasonably complete ISO C++ implementation appears?

For some reasonable definition of "reasonably complete" this happened
a few months ago.

I can do most of what I need to do using EDG-based compilers, GNU's EGCS,
MS and Borland compilers, and Sun just released their 5.0 which is claimed
to be pretty close.

The remaining different features still vary a bit. They tend to be something
like partial specialization, "export", or Koenig lookup. However, there
clearly is rapid convergence in the latest releases.

 - Bjarne
Bjarne Stroustrup - http://www.research.att.com/~bs



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Steve Clamage wrote in message <7bugfd$nin$1@engnews1.eng.sun.com>...

>The C++ language standard is not a spec designed for manufacture,
>nor should it be. It provides a definition of the language, not
>a specification for a compiler.


One could draw quite the opposite conclusion by reading the first sentence
of 1.1 Scope.




[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

 +---- bs@research.att.com wrote (8 Mar 1999 18:52:11 GMT):
 | For some reasonable definition of "reasonably complete" this happened
 | a few months ago.

Someone should update the FAQ.

 | I can do most of what I need to do using EDG-based compilers, GNU's EGCS,
 | MS and Borland compilers, and Sun just released their 5.0 which is claimed
 | to be pretty close.
 |
 | The remaining different features still vary a bit. They tend to be something
 | like partial specialization, "export", or Koenig lookup. However, there
 | clearly is rapid convergence in the latest releases.
 +----

So there aren't any major pieces that are commonly left
unimplemented, or partially implemented?  Are the variant bits
confined to little used areas?

--
Gary Johnson     gjohnson@season.com
Privacy on the net is still illegal.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Biju Thomas wrote in message <36E3ED63.631E0430@ibm.net>...
>Andrei Alexandrescu wrote:
>>
>> John Lilley has taken the grammar of a draft standard and expressed it in
>> pccts' EBNF format, together with symbol table management. He's put
>> everything in the public domain.
>
>Can you give the location where this C++ grammar is available?


Here it is:

http://www.empathy.com/pccts/

I compel you all to take a look at it. It's an excellent starting
point for an open-source "C++ proofer".
Cheers,

Andrei



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.114a5752cd41bf19989b66@netnews.worldnet.att.net>,
  Tony@ask.me (Tony) wrote:
> In article <36DF3EB5.3CE14A6D@wizard.net>, kuyper@wizard.net says...
> > From the anti-trust point of view, standardization is comparable to
> > price-fixing. Price-fixing inhibits price-based competition;
> > standardization inhibits feature-based competition.
>
> Standards should probably be well accepted practices and lowest-common-
> denominator-like things where there is no question as to their motives.
> From that foundational base, there is plenty of room for value-added
> features and innovation.  What is sad, as you allude to, is the situation
> where "standards" (de facto perhaps) eliminate choice of the user and
> oppress potential developers (e.g. the GUI situation).

I think that what Tony is getting at, is that intent matters. If
ANSI/ISO didn't have exemptions in federal law, then someone could
conceivably sue for anti-trust. However, such a lawsuit would
surely fail if ISO could demonstrate that

  1) The organization allowed anyone to become a member for a
     relatively small investment
  2) The organization had no ability nor desire to force anyone
     to follow the standards
  3) Even those companies that do follow the standards, are not
     obligated to leave out additional features

These things would all be easy to prove, if we assume that ANSI
and ISO behave the way that they do today, and if we also assume
that any jury would consist of reasonable people who want to be
fair. The latter assumption has had glaring exceptions in recent
memory, but overall I expect that it is true much more often than
not.

Then again, I'm not a lawyer -- so please don't assume that I
know what I'm talking about.  :^]

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.114b5f4319abb5b98973c@news.mindspring.com>,
  brahms@mindspring.com (Stan Brown) wrote:
> In this next-to-last year of the millennium, Stephen.Clamage@sun.com
> (Steve Clamage) wrote in article <7brork$lki$1@engnews1.eng.sun.com> in
> comp.std.c++:
> >Now we are back to assuming that you can create an implementation
> >known to be bug-free for something as complex as C++ in a reasonable
> >time frame. If that were possible, we'd have complete and correct
> >compilers now.
>
> Not necessarily (see below).
>
> >I claim it isn't possible in any time frame of interest.
>
> Literally true, Steve, but I think it's beside the point.

I don't suspect that the timeframe for competent people working on an
ANSI or ISO standard, would be all that different from the timeframe
for compiler makers who care about release cycles. The only way this
could be negative is if the compiler makers waited for the reference
compiler *AND THEN* started working on their own version.

> On freedom from bugs:
> ====================
>
> The Standard itself isn't bug free, but that doesn't prevent it from
> being useful. Isn't that's why there's a defect-reporting procedure?
>
> I think none of us has ever seen a bug-free program of any size,

Of course we have. Write a trivial program, such as one that reports
the sum of all numbers on the command line. If you can't get that
bug free, then your definition of "bug" is too opressive.

> but we
> use them and try to work around the bugs. As software writers, surely we
> know by now that correctness is a sort of mathematical limit we can
> approach as closely as we like, given sufficient time and money, but
> never quite reach.

Tell that to NASA. Reaching it is expensive, but possible.

> The 90-10 rule tells us that the effort to eliminate
> bug number N tends to increase nonlinearly with N.

Which doesn't make N infinite! Also, this is a rule of thumb that has
proven very useful, but has never been subjected to rigid proofs. For
instance, the numbers "90" and "10" are surely approximations, differing
(at least slightly!) from one project to the next.

> A reference implementation could not be bug free any more than any other
> large effort.

If a reference implementation had a significant number of non-trivial
bugs, it would be useless.

> But, as (I think) Francis observed, it would not have to be
> efficient. Eliminating the optimizer phase would remove a big chunk of
> effort (and playing field for bugs).

So we have the need to be totally bug-free, but not the need for more
than trivial optimizations. Perhaps these things cancel each other out
in terms of implementation time.

> You suggest that if it were possible to create a bug-free implementation
> in reasonable time we'd have one by now. I think there's a hidden
> assumption there: that compiler makers have it as their priority but have
> failed.

Whereas, clearly at least some compiler makers define "good enough" as
"anything that doesn't hurt sales too much."

> Commercial implementations don't seem to care much about standard
> compliance.

Or about provable correctness, or even incidental correctness. I doubt
that my cynicism applies to every major manufacturer, but it certainly
applies to most of the big ones.

> The writers of the reference implementation would care only
> about standard compliance -- no jazzy interface, no "foundation classes".
>
> I don't have wide experience, but I think it's pretty clear that at least
> for Brand M (which I do know fairly well), standard compliance is not a
> priority -- at least not when compared to a jazzy interface and frequent
> releases.

Brand M is improving. They were one of the first to automate testing,
which allows them to find 100% of regression bugs. However, I have no
evidence that they even know what it means to prove that an algorithm
is correct, other than to give out 50,000 beta copies and wait for
bug reports.

> (If correctness were a priority for Brand M, the compiler would
> compile the supplied header files in "ISO/ANSI mode" without diagnostics,
> and it doesn't.) And for pretty much any compiler, marketplace acceptance
> *must* be a higher priority than anything else.

True. However, when one manufacturer becomes big enough to guide
marketplace acceptance, the rules get less strict. IBM was once
in that position, and Brand M is today.

> On the amount of effort required:
> ================================
>
> I readily admit that I've never written a compiler, so I probably
> underestimate the difficulty. But I think it's important not to
> OVERestimate it. A reference implementation would not have to have the
> baggage of commercial implementations. It would not have to worry about
> compile speed or speed of generated code.

However, even 3-4 minor bugs would make it useless, whereas most of us
have first-hand evidence that 8-10 major bugs won't slow down Brand M.

> In fact, perhaps a reference implementation would not even need to
> generate code.

A "reference implementation" that does not generate code would amount
to little more than a "lint" program. Programmers need to know that
different compilers will not only accept the same syntax, but will
provide the same semantics (except where not mandated in the standards,
such as the order of evaluations). In other words, Brand M must not
only accept all programs that the reference implementation does, but it
must also execute in the same way. If the reference implementation does
not execute, then this becomes impossible.

> Maybe it could generate C code;

That would probably be acceptable, if the C code was low-level enough.
Again, the goal is to demonstrate what the program does when executed.
If we translate C++ features into equivalent C features, then we will
need a reference implementation of C as well!

> maybe it would not even do
> that, but just say whether a given program was conforming.

Useless. See above.

> Of course, it
> would certainly be most useful if it generated *something* that could
> ultimately be run, so that we could test the results of other
> implementations against the reference.

Yes.

> Also, a reference implementation could be created by people who are not
> thrashing out the language definition, though doubtless there would be
> overlap. But I wonder if there need be much overlap. For instance,
> probably all of us can write programs and all can debug programs. But
> some are much better at one than the other. Debugging and original
> programming are two different skills. In the same way, I wonder if
> "language-lawyering" might be a different skill from compiler-writing, so
> that having people create a reference implementation in parallel with the
> language definition would be possible without siphoning too much talent
> from writing the language definition.
>
> Final thought:
> =============
>
> I think a reference implementation would be wonderful to have.
>
> Look at how many people try to discover whether something is legal C++ by
> trying it in different compilers. It's easy to do and gives a clear
> answer. The problem, as we know, is that though the answer is clear it
> may not be correct. The reference implementation's answer would be clear
> and correct (modulo any defect reports, which would also be clear and
> unambiguous).

Good point, but this also gives the reference implementation an extra
requirement that commercial versions would not have: *ALL* invalid
programs *MUST* fail to compile or execute correctly. i.e. we had better
make sure that a program which casts a void* to a long and back again
doesn't run without any diagnostic errors. Otherwise, we will have at
least some programmers using the reference implementation to prove that
this is legal, and then demand that the commercial versions do so as well.

> My reply address is correct as is. The courtesy of providing a correct
> reply address is more important to me than time spent deleting spam.

How nice. Personally, I find the 30 minutes I spend deleting spam is already
enough; I don't want to invite any more. Therefore:  AllanW@my-dejanews.com
is a "Spam Magnet" -- never read.  Please reply in USENET only, sorry.
http://www.dejanews.com/[ST_rn=ps]/dnquery.xp?ST=PS&authors=allanw%40my-dejan
ews .com

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

gjohnson@dream.season.com (Reality is a point of view) writes:

> +---- bs@research.att.com wrote (8 Mar 1999 18:52:11 GMT):
> | For some reasonable definition of "reasonably complete" this happened
> | a few months ago.

>Someone should update the FAQ.

> | I can do most of what I need to do using EDG-based compilers, GNU's EGCS,
> | MS and Borland compilers, and Sun just released their 5.0 which is claimed
> | to be pretty close.
> |
> | The remaining different features still vary a bit. They tend to be something
> | like partial specialization, "export", or Koenig lookup. However, there
> | clearly is rapid convergence in the latest releases.
> +----

>So there aren't any major pieces that are commonly left
>unimplemented, or partially implemented?  Are the variant bits
>confined to little used areas?

I don't know how to quantify "reasonably complete," or "major
pieces", or "little-used" areas.

I made an attempt one time, characterizing a compiler as being
nearly complete. My argument was based on the page count in the
standard: less than 5% of the page count was devoted to features
that weren't implemented.

Someone objected, saying that, for example, the basic types take up
very few pages, so page count was not a good criterion. We could
argue that point indefinitely, but in the end, what matters to you is
whether the features your code depends on are available and reliable.

I don't know of any compiler that supports the full semantics
of exported templates, or supports template template parameters.
Your code won't depend on these features, because they are
not available!

Not many compilers support template members of classes. The new
syntax for calling template functions -- f<T,U>(a,b,c) -- is not
widely supported. If you depend on those features, your choice
of compilers will be limited.

With some notable exceptions, most recent compilers do a good job
of supporting the C++ standard. The exceptions are those vendors
who have serious compatibility issues that compliance would break,
or who have other issues that are more important to them than
standards compliance.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

AllanW@my-dejanews.com wrote:
>
> In article <MPG.114a5752cd41bf19989b66@netnews.worldnet.att.net>,
>   Tony@ask.me (Tony) wrote:
> > In article <36DF3EB5.3CE14A6D@wizard.net>, kuyper@wizard.net says...
> > > From the anti-trust point of view, standardization is comparable to
> > > price-fixing. Price-fixing inhibits price-based competition;
> > > standardization inhibits feature-based competition.
> >
> > Standards should probably be well accepted practices and lowest-common-
> > denominator-like things where there is no question as to their motives.
> > From that foundational base, there is plenty of room for value-added
> > features and innovation.  What is sad, as you allude to, is the situation
> > where "standards" (de facto perhaps) eliminate choice of the user and
> > oppress potential developers (e.g. the GUI situation).
>
> I think that what Tony is getting at, is that intent matters. If
> ANSI/ISO didn't have exemptions in federal law, then someone could
> conceivably sue for anti-trust. However, such a lawsuit would
> surely fail if ISO could demonstrate that
>
>   1) The organization allowed anyone to become a member for a
>      relatively small investment

I'm no lawyer, but I can make an analogy with price-fixing laws. The
more people that participate in a price-fixing agreement, the more
effective it is, and the less legal it is. Having no membership
restrictions isn't a defense. In fact, some of the most effect price
fixing agreements have been the ones that have so many participants that
they've been able to suborn the government into making the price-fixing
a matter of law. Consider milk price supports in the US, for example.

>   2) The organization had no ability nor desire to force anyone
>      to follow the standards
>   3) Even those companies that do follow the standards, are not
>      obligated to leave out additional features

Feature-based competition runs both ways - dropping features to lower
costs, or adding more features to increase value. Standardization, if it
is to have any meaning at all, inhibits the former kind of competition.
It also inherently inhibits the taking of incompatible choices, which
might be better ones. That's why it's a dangerous to allow standards to
be set by any process that gives individual companies too much direct
influence.

The original idea on this thread was to produce a standard outside of
ISO's normal slow, complicated procedures. The problem with that is that
whatever power ISO does have, is the result of earning a reputation as
an organization that produces well-considered standards with
international participation in the process. I'm not saying that every
ISO procedure is well-justified; I don't even know enough about the
detailed procedures to comment on them. I am saying that slow, carefully
considered consensus building is a good idea in this context, and that
this proposal seems motivated by a desire to avoid precisely those
features of the ISO process.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7c0s08$576$1@engnews1.eng.sun.com>,
  Stephen.Clamage@sun.com (Steve Clamage) wrote:
> The main hurdle remains: who will do the work? It's easy for
> non-participants to say that "someone" should do it. But
> non-participants have already shown that they themselves
> are unwilling or unable to contribute.

For some of us, the unable is much bigger than the unwilling.

Let me plead "poor programmer" for a while. I am perfectly willing to
assist in any reasonable way with the standardization effort. Unfortunately,
with no corporate backing, "reasonable" means that I could not afford to
pay more than a few dozen dollars, and travel more than 100 miles one-way
would be impractical. Meetings held during business hours, while not
impossible, are difficult to manage even if they are within the 100-mile
limit.

Anything I can do from home, up to 20- or 30-hours per week, I will
gladly do. That is one reason that I begged to have draft standards
made public when they were still a work in process.  I realize and
respect the reasons why there weren't more public drafts, but it
did also limit my involvement quite a bit.

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 09 Mar 99 01:57:20 GMT, AllanW@my-dejanews.com wrote:
[...]
>A "reference implementation" that does not generate code would amount
>to little more than a "lint" program.

Surely you jest. Writing a conforming front end for a C++ compiler is
a very large amount of work. In all cases I know of, the task of
writing a front end (for languages much simpler than C++) was
substantially larger than writing the back end, optimizers &c.

---------------------------------------------
Ziv Caspi
zivca@netvision.net.il
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Steve Clamage <Stephen.Clamage@sun.com> wrote:

: It seems to me the best chance for a reference implementation
: along the lines I have suggested would have been g++ or egcs.
: The much-vaunted "bazaar" development model should have
: produced a reference implementation. A number of contributors
: to g++/egcs are long-time C++ committee members. So why didn't
: it happen? Does anyone from that bazaar want to comment?
: (I think the answer is the difficulty of producing a C++
: implementation -- the missing features are still being worked on.)

g++ has never followed the bazaar development model. g++ development
was stalled for about 2 years partially because of this. That's why
egcs project came to life. egcs is about 1.5 years old, yet it
has done more in this time then SunPro did in the last 5 years with
the much-vaunted non-bazaar development model.

Oleg.
--
Life is a sexually transmitted, 100% lethal disease.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:

>In article <7bks6a$o9u$1@engnews1.eng.sun.com>, stephen.clamage@sun.com
>says...
>>
>> Tony@ask.me (Tony) writes:
>>
>> >My question is though: Can the standards participants be persuaded
>> >(required?) to produce a _reference_implementation_?  Has that been
>> >talked about at all?  I believe it would lead to better (a practicianer's
>> >std vs. a computer scientist's std, if you will) standards.
>>
>> Currently, a number of implementors are working at top speed to
>> produce implementations that are as complete and correct as they
>> can possibly make them.
>>
>> So far, no one has succeeded in making one known to be complete
>> and correct. Even for just the parts of the standard that are
>> implemented, after the implementation passes all known test
>> suites, bug reports still come in from users in the field.

>What do the above 2 paragraphs have to do with the standards _committee_
>producing a reference implementation?

>> Where would the reference implementation come from, and how
>> would we know it is correct?

>I thought my question obviously stated that those who are "throwing the
>requirements over the wall to engineering/production" would assume the
>responsibility/rise to the occasion.  Have you heard of concurrent
>engineering?  There is a definite point in time that occurs on a project
>after which, if you persist in separating specification from
>implementation, much potentially useful information gets lost.  (And
>probably a whole host of other unbeneficial effects occur too).

Aha!  We were using the term "reference implementation" to mean
two entirely different things.

I normally associate that term with an implementation that defines
the language (or whatever is being standardized). I've had the
unhappy experience of having to produce a compiler for a language
that had an inadequate written spec and a reference implementation
whose behavior we had to duplicate.

I see now that you meant a concurrent development that would
expose problems with the specification. I agree that is a
wonderful idea.

In theory, the standards committee is supposed to standardize
existing practice, and not invent a whole new language. In
practice, the C++ committee came closer to the latter than to
the former. ("In theory, theory and practice are the same;
in practice they are different.")

The members of the committee who were implementers worked hard
to try to keep up with the emerging specifications, but it is
easier to vote in new requirements than it is to implement them.

I hope than in a future round of standardization we will be
voting only on features that have already been in use.

--
Steve Clamage, stephen.clamage@sun.com
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <zaBKQSB$jX32EwPk@robinton.demon.co.uk>,
  Francis Glassborow <francisG@robinton.demon.co.uk> wrote:
> In article <7baoaf$f0b$1@remarQ.com>, Joe Buck <jbuck@synopsys.com>
> writes
> >Yes, the standards organizations need a funding mechanism.
> >I'm not convinced that the current method is the only option.
>
> So propose an alternative.

Govermental sponsorship? [Hey, if the U.S. Federal Government pays
high school students to burn U.S. flags in the name of art, or pays
tobacco farmers NOT to grow tobacco to help keep the price high,
then it can certainly pay it's fair share of costs to develop an
international standard for a programming language!]

Royalties? [All C++ compilers that claim to be conforming -- weather
it's true or not -- must pay 1% of the retail sales price, or $1,000
per year, whichever is more, provided that the party of the first
part makes, or has made, or will make in a timely manner, the source
code of the software available to the party of the second part, unless
that source code falls under the provision of section B, in which
case...]

Free advertising through thinly disguised Corporate Sponsorship [The
C++ standard was made possible by a grant from Microsoft, makers of
Office97. Mention the C++ standard and receive 5% off of your next
Microsoft purchase of $1,000 or more. For the math-impaired (the
majority of our customers), that's worth at least $50. And if you
act within the next two days, you also get...]

Bake sale? [$2 per slice. But this isn't just cake, it's cake++]

Shareware? [This standard is not free, it's shareware. If you like
it and keep using it, you should send $18 to ISO. Registered users
will receive the table of contents for FREE...]

PBS-style pledge drive? [In a moment, we'll have another chapter
of the C++ standard. But first, let's see how well the phones are
doing...]

TV public service announcements [These programmers don't have to
suffer! For just 5 cents per day, you can keep 50 programmers
working on the next standard...]

Cut corners [Just slap the word "Java" on it; if nobody notices, we
can sell one book with two titles and collect twice as much money!]

Yearly telethons [Ha ha, thank you Robin Williams. Okay, I have a
challenge to all ADA programmers, for every dollar you donate in
the next 12 minutes Bee-jarne will donate an extra 50 cents.
Sorry, I'm told that's pronounced Be-ar-neh. Okay, now it's time
for Vanna White to show us the big board...]

Like the Hollywood entertainment business, start holding
self-congratulatory "awards" ceremonies. ["Best suggestion for
syntax related to namespaces and function calls," the winner is:
Andrew Koenig, for Koenig name lookup! <applause>]

Funding through the Psychic Network [members already know about
this...]

Blackmail [For $1,000,000, we'll change the specs on auto_ptr<>
again, but this time we'll make it so that it matches the
implementation you're already using. Better act soon, because
Sun is already trying to buy it just to spite Microsoft...]

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Barry Margolin wrote:
>
> In article <36DF3EB5.3CE14A6D@wizard.net>,
> James Kuyper  <kuyper@wizard.net> wrote:
> >I think most people reading this newsgroup are likely to agree that
> >standardization is a good idea, but there needs to be a barrier
> >preventing it from being misused. For example, consider what a
> >Microsoft-dominated committee would have done. In the opposite
> >direction, consider how Microsoft feels about Sun's sponsorship of the
> >work on a Java standard.
>
> ANSI and ISO rules are supposed to prevent things like a
> "Microsoft-dominated committee".  In ANSI committees, the rule is "one
> company, one vote"; in ISO it's "one country, one vote".

Yes, that's why it's reasonable for anti-trust law to have an exemption
for such organizations.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36DF3EB5.3CE14A6D@wizard.net>, kuyper@wizard.net says...
> From the anti-trust point of view, standardization is comparable to
> price-fixing. Price-fixing inhibits price-based competition;
> standardization inhibits feature-based competition.

Standards should probably be well accepted practices and lowest-common-
denominator-like things where there is no question as to their motives.

In article <7bpf0j$abf$1@nnrp1.dejanews.com>, AllanW@my-dejanews.com
says...
>
> In article <zaBKQSB$jX32EwPk@robinton.demon.co.uk>,
>   Francis Glassborow <francisG@robinton.demon.co.uk> wrote:
> > In article <7baoaf$f0b$1@remarQ.com>, Joe Buck <jbuck@synopsys.com>
> > writes
> > >Yes, the standards organizations need a funding mechanism.
> > >I'm not convinced that the current method is the only option.
> >
> > So propose an alternative.
>
> Govermental sponsorship? [Hey, if the U.S. Federal Government pays
> high school students to burn U.S. flags in the name of art, or pays
> tobacco farmers NOT to grow tobacco to help keep the price high,
> then it can certainly pay it's fair share of costs to develop an
> international standard for a programming language!]

Did someone say something about _the_people_ owning the infrastructures?
Heck if we eliminate political opinion polls and tabloid politics, there
would be plenty of the green stuff for _productive_ purposes like those.

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bpf0j$abf$1@nnrp1.dejanews.com>, AllanW@my-dejanews.com
writes
>Royalties? [All C++ compilers that claim to be conforming -- weather
>it's true or not -- must pay 1% of the retail sales price, or $1,000
>per year, whichever is more, provided that the party of the first
>part makes, or has made, or will make in a timely manner, the source
>code of the software available to the party of the second part, unless
>that source code falls under the provision of section B, in which
>case...]

Actually, I have always liked this option.  I would extend it to include
all libraries and other tools (you would have to because otherwise
compiler producers will claim that the compiler itself is a tiny
fraction of the cost of the whole.)

In addition compilers should then be delivered with at least a soft copy
of the relevant standard(s) (the idea should extend to more than C++).

The payment should be made to the standards body of the country in which
the distribution subsidiary is operating.

Now the main problem is that we would need to lobby our governments to
provide the relevant legal infrastructure.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bp4q5$nub$1@engnews1.eng.sun.com>, Steve Clamage
<Stephen.Clamage@sun.com> writes
>I sincerely hope that Francis meant "draft".

Ah!! A cross Atlantic problem.  One meaning of 'graft' in English is
'hard work'

>
>I am NOT a crook! :-)

Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36DEE01A.F8081516@technologist.com>,
dtribble@technologist.com says...
> There's also the small problem of time that requiring a reference
> implementation would cost.  If the ISO committee had to provide
> a working implementation (no matter how inefficient), how much
> time would people be willing to wait in order to delay the
> acceptance of the standard while the implementation was being
> written?

What I was trying to promote is the synergy of the approach where a
mutual dependency between specification and implementation would produce
something greater than the sum of the parts.

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36DEE01A.F8081516@technologist.com>,
dtribble@technologist.com says...
> Bear in mind also that the implementation would probably
> need to be written by committee (or groups of subcommittees),
> simply because of the size and scope of the language and library,
> which would add further delays when it came time to integrate the
> pieces.

>
> The ISO standard was already delayed about two years (while the
> STL was being rolled into it); would anyone have wanted to wait even
> longer for final acceptance?

So why create a mononlithic standard anyway?  (Analogy: Mach and BSD UNIX
kernels).

Tony



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bp4m2$nhh$1@engnews1.eng.sun.com>, clamage@eng.sun.com
says...
> Aha!  We were using the term "reference implementation" to mean
> two entirely different things.
>
> I normally associate that term with an implementation that defines
> the language (or whatever is being standardized).
>
> I see now that you meant a concurrent development that would
> expose problems with the specification.

Well I wasn't emphasizing one as more important than the other.

In theory, the two (specification and implementation) would
_evolve_ concurrently and should be in complete agreement at the end of
the project.  The influence of the implementation on the specification, I
believe, would result in a standard of higher quality.

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <eOMyPIAoF932EwrY@robinton.demon.co.uk>,
francis@robinton.demon.co.uk says...
> In article <MPG.1148da60366d462f989b53@netnews.worldnet.att.net>, Tony
> <Tony@ask.me> writes
> >That should be easily doable just by using the money that is
> >currently being used to congregate (someone said 8k per person per
> >trip!).  Have all interactions occur on the net, dropping expenses to
> >almost nil.  :)
>
> Sorry (yes I see the smiley) but that really does not work people still
> need to meet (and much of the graft is already done over the Internet).
> Anyway how does that cut the cost of time.  The real cost is not the
> airfares and hotel bills but in the work time spent away from the
> employers product.

My guess is that like any committee project, too much is being discussed
in too much detail.  Is the project management of these events any good?

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:

>In article <eOMyPIAoF932EwrY@robinton.demon.co.uk>,
>francis@robinton.demon.co.uk says...
>> In article <MPG.1148da60366d462f989b53@netnews.worldnet.att.net>, Tony
>> <Tony@ask.me> writes
>> >That should be easily doable just by using the money that is
>> >currently being used to congregate (someone said 8k per person per
>> >trip!).  Have all interactions occur on the net, dropping expenses to
>> >almost nil.  :)
>>
>> Sorry (yes I see the smiley) but that really does not work people still
>> need to meet (and much of the graft is already done over the Internet).
>> Anyway how does that cut the cost of time.  The real cost is not the
>> airfares and hotel bills but in the work time spent away from the
>> employers product.

>My guess is that like any committee project, too much is being discussed
>in too much detail.

If you don't discuss everything in great detail, you don't produce
a standard of acceptable quality. This is a language standard.
Every sentence matters. Every language detail must be covered.

Reasonable people can differ reasonably on how features should
be designed and defined. If you don't get consensus, you don't
get a standard that will be used. Case in point:  Most people
don't know that there is an ANSI standard for BASIC. It defines
a language that no one would bother to implement or use. It
bears little resemblance to popular forms of BASIC that do
get used.

>Is the project management of these events any good?

This is an all-volunteer effort. If you are volunteering to take
over the project and produce a better result faster, I will be
happy, nay eager, to resign as chair in your favor.

If you haven't been part of a standards process, you don't appreciate
where many of the difficulties lie. It isn't anything like producing a
commercial product.

--
Steve Clamage, stephen.clamage@sun.com
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

>The trouble is that the interests of the organization and the interests
>of its volunteer members do not coincide.

The members volunteered, joined, and participated with full understanding
that the fruits of their labor would become the intellectual property of the
organization. Their expectations should reflect nothing more than this.

>The development of the standard
>was primarily funded by the volunteer members, not by ISO.  I think most
>of the volunteer members that developed this standard would be happy to
>see it freely available.


Irrespective of the relative monetary contributions of the organization vs
the members, the organization does incur expenses in the maintenance of
itself and the distribution of the document. How else to recover those
expenses than to charge a nominal fee for copies of the document? In this
particular enterprise the document is their only asset.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

>There's also the small problem of time that requiring a reference
>implementation would cost.  If the ISO committee had to provide
>a working implementation (no matter how inefficient), how much
>time would people be willing to wait in order to delay the
>acceptance of the standard while the implementation was being
>written?  Bear in mind also that the implementation would probably
>need to be written by committee (or groups of subcommittees),
>simply because of the size and scope of the language and library,
>which would add further delays when it came time to integrate the
>pieces.

Time notwithstanding, it is not altogether unreasonable (in my opinion) to
require that committees that mandate programming language standards (that do
not merely codify prior art) be required to prove their innovations with a
working prototype translator before the standard may be formally approved.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:

>In article <7bp4m2$nhh$1@engnews1.eng.sun.com>, clamage@eng.sun.com
>says...
>> Aha!  We were using the term "reference implementation" to mean
>> two entirely different things.
>>
>> I normally associate that term with an implementation that defines
>> the language (or whatever is being standardized).
>>
>> I see now that you meant a concurrent development that would
>> expose problems with the specification.

>Well I wasn't emphasizing one as more important than the other.

>In theory, the two (specification and implementation) would
>_evolve_ concurrently and should be in complete agreement at the end of
>the project.  The influence of the implementation on the specification, I
>believe, would result in a standard of higher quality.

Now we are back to assuming that you can create an implementation
known to be bug-free for something as complex as C++ in a reasonable
time frame. If that were possible, we'd have complete and correct
compilers now.

I claim it isn't possible in any time frame of interest.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

"Al Stevens" <alstevens@midifitz.com> writes:

>>The trouble is that the interests of the organization and the interests
>>of its volunteer members do not coincide.
>
>The members volunteered, joined, and participated with full understanding
>that the fruits of their labor would become the intellectual property of the
>organization. Their expectations should reflect nothing more than this.

Certainly.  But their expectations may fall well short of their desires!

The current situation where standards are not freely distributable is
highly undesirable.  We participate, despite knowing this, because the
alternative of not participating is even worse -- the result if we did
not participate would be a worse standard, and it still wouldn't be
freely distributable.  The fact that we participate does not imply that
we don't yearn for something better.

As individuals (or even individual companies) we can't do much to improve
the situation.  But as a group, we have a lot more power.

The Ada standard is freely available, apparently someone said
the Lisp (or was it Scheme) standard is freely available,
and many non-ISO de facto standards such as the Haskell report
are freely available.  So clearly it is possible to create
standards, even ISO standards, which are freely distributable.

>>The development of the standard
>>was primarily funded by the volunteer members, not by ISO.  I think most
>>of the volunteer members that developed this standard would be happy to
>>see it freely available.
>
>Irrespective of the relative monetary contributions of the organization vs
>the members, the organization does incur expenses in the maintenance of
>itself and the distribution of the document. How else to recover those
>expenses than to charge a nominal fee for copies of the document? In this
>particular enterprise the document is their only asset.

By all means, let ISO charge a fee for distribution of the document.
But let the document be freely copyable, and let ISO compete with
anyone else who wants to distribute copies for a fee, just as
e.g. RedHat and CheapBytes compete.  If ISO can't compete on that
basis, then arguably they shouldn't be in the document distribution
business -- let ISO rubber-stamp it, but someone else can distribute it.

That only leaves the expenses that ISO incurs in the maintenance of itself.
But if the a different organization is creating the standards, and a
different organization is distributing them, then what do we need ISO for?

--
Fergus Henderson <fjh@cs.mu.oz.au>  |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>  |  of excellence is a lethal habit"
PGP: finger fjh@128.250.37.3        |     -- the last words of T. S. Garp.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <F850xy.6po@research.att.com>, ark@research.att.com says...
> In article <MPG.1148da60366d462f989b53@netnews.worldnet.att.net>,
> Tony <Tony@ask.me> wrote:
>
> > > Sure thing -- just offer to pay them enough to make it worth their while.
> > > A few million bucks should do the trick.
>
> > That should be easily doable just by using the money that is
> > currently being used to congregate (someone said 8k per person per
> > trip!).  Have all interactions occur on the net, dropping expenses to
> > almost nil.  :)
>
> The trouble with statements of the form ``that should be easily doable''
> is that it doesn't mention who would do it, or why they would be
> interested in doing so.

You obviously didn't read the whole post which read like: "save in one
area, apply to another".

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In this next-to-last year of the millennium, Stephen.Clamage@sun.com
(Steve Clamage) wrote in article <7brork$lki$1@engnews1.eng.sun.com> in
comp.std.c++:
>Now we are back to assuming that you can create an implementation
>known to be bug-free for something as complex as C++ in a reasonable
>time frame. If that were possible, we'd have complete and correct
>compilers now.

Not necessarily (see below).

>I claim it isn't possible in any time frame of interest.

Literally true, Steve, but I think it's beside the point.

On freedom from bugs:
====================

The Standard itself isn't bug free, but that doesn't prevent it from
being useful. Isn't that's why there's a defect-reporting procedure?

I think none of us has ever seen a bug-free program of any size, but we
use them and try to work around the bugs. As software writers, surely we
know by now that correctness is a sort of mathematical limit we can
approach as closely as we like, given sufficient time and money, but
never quite reach. The 90-10 rule tells us that the effort to eliminate
bug number N tends to increase nonlinearly with N.

A reference implementation could not be bug free any more than any other
large effort. But, as (I think) Francis observed, it would not have to be
efficient. Eliminating the optimizer phase would remove a big chunk of
effort (and playing field for bugs).

You suggest that if it were possible to create a bug-free implementation
in reasonable time we'd have one by now. I think there's a hidden
assumption there: that compiler makers have it as their priority but have
failed.

Commercial implementations don't seem to care much about standard
compliance. The writers of the reference implementation would care only
about standard compliance -- no jazzy interface, no "foundation classes".

I don't have wide experience, but I think it's pretty clear that at least
for Brand M (which I do know fairly well), standard compliance is not a
priority -- at least not when compared to a jazzy interface and frequent
releases. (If correctness were a priority for Brand M, the compiler would
compile the supplied header files in "ISO/ANSI mode" without diagnostics,
and it doesn't.) And for pretty much any compiler, marketplace acceptance
*must* be a higher priority than anything else.

On the amount of effort required:
================================

I readily admit that I've never written a compiler, so I probably
underestimate the difficulty. But I think it's important not to
OVERestimate it. A reference implementation would not have to have the
baggage of commercial implementations. It would not have to worry about
compile speed or speed of generated code.

In fact, perhaps a reference implementation would not even need to
generate code. Maybe it could generate C code; maybe it would not even do
that, but just say whether a given program was conforming. Of course, it
would certainly be most useful if it generated *something* that could
ultimately be run, so that we could test the results of other
implementations against the reference.

Also, a reference implementation could be created by people who are not
thrashing out the language definition, though doubtless there would be
overlap. But I wonder if there need be much overlap. For instance,
probably all of us can write programs and all can debug programs. But
some are much better at one than the other. Debugging and original
programming are two different skills. In the same way, I wonder if
"language-lawyering" might be a different skill from compiler-writing, so
that having people create a reference implementation in parallel with the
language definition would be possible without siphoning too much talent
from writing the language definition.

Final thought:
=============

I think a reference implementation would be wonderful to have.

Look at how many people try to discover whether something is legal C++ by
trying it in different compilers. It's easy to do and gives a clear
answer. The problem, as we know, is that though the answer is clear it
may not be correct. The reference implementation's answer would be clear
and correct (modulo any defect reports, which would also be clear and
unambiguous).

--
Stan Brown, Oak Road Systems, Cleveland, Ohio, USA
                                    http://www.mindspring.com/~brahms/
My reply address is correct as is. The courtesy of providing a correct
reply address is more important to me than time spent deleting spam.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7brork$lki$1@engnews1.eng.sun.com>, Stephen.Clamage@sun.com
says...
>
> Tony@ask.me (Tony) writes:
>
> >In article <7bp4m2$nhh$1@engnews1.eng.sun.com>, clamage@eng.sun.com
> >says...
> >> Aha!  We were using the term "reference implementation" to mean
> >> two entirely different things.
> >>
> >> I normally associate that term with an implementation that defines
> >> the language (or whatever is being standardized).
> >>
> >> I see now that you meant a concurrent development that would
> >> expose problems with the specification.
>
> >Well I wasn't emphasizing one as more important than the other.
>
> >In theory, the two (specification and implementation) would
> >_evolve_ concurrently and should be in complete agreement at the end of
> >the project.  The influence of the implementation on the specification, I
> >believe, would result in a standard of higher quality.
>
> Now we are back to assuming that you can create an implementation
> known to be bug-free for something as complex as C++ in a reasonable
> time frame. If that were possible, we'd have complete and correct
> compilers now.
>
> I claim it isn't possible in any time frame of interest.

Then to me that sounds like the spec is not adequately "designed for
manufacture" and needs more work.  Just the kind of thing concurrent
engineering should prevent.

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.114a583e67633f36989b67@netnews.worldnet.att.net>, Tony
<Tony@ask.me> writes
>My guess is that like any committee project, too much is being discussed
>in too much detail.  Is the project management of these events any good?

A requirement for a standard is that those involved get as close to
consensus as possible.  For something as important as a standard that is
actually important.  You probably have little idea as to how often
serious errors have been prevented by the sharp eyes of a single person.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony:
>>My guess is that like any committee project, too much is being discussed
>>in too much detail.

Instead of guessing, why not attend and find out? You would be
welcome.

Steve:
>If you don't discuss everything in great detail, you don't produce
>a standard of acceptable quality. This is a language standard.
>Every sentence matters. Every language detail must be covered.

Indeed. You need "language lawyers" on programming language standards
committees for the same reason you need lawyers on legal teams: They
bring a thorough and meticulous understanding of the issues,
interactions, and interrelationships of various features/laws, and
they protect your behind. Like lawyers, committee members specialize:
I participate in the library working group (there used to be several
of those, covering various parts of the library); others participate
on the core language working group (there used to be several of those
too); now we are in the process of forming a performance/usability
working group, and many people are actively involved in getting that
off the ground.

If you're sometimes surprised or dismayed at some of the details of
how some C++ features interact, think how much worse it would have
been without extensive detailed discussion among experts who
specialize in things like iostreams, STL, templates, exceptions,
locales, etc.

Tony:
>>Is the project management of these events any good?

Come and find out! :) We're meeting in Dublin next month. See this
newsgroup's FAQ to learn how you can get involved.

Herb


---
Herb Sutter (mailto:hsutter@peerdirect.com)

PeerDirect Inc.     2695 North Sheridan Way, Suite 150
www.peerdirect.com  Mississauga Ontario Canada L5K 2N6


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Bjarne Stroustrup (bs@research.att.com) wrote:
: Some of us decided that we could try to reform the UN (that officially
: controls the ISO) and all the national standards bodies (ANSI, AFNOR, BSI,
: DIN, ...) or we could produce a reasonably timely standard; however we
: we couldn't do both.

This is unproductive. Few, if any of us, are claiming that the creation of
the standard under the auspices of ISO was a mistake. The question now is
what can we do to make the standard more widely used and more easily used
for the whole community.

: Some of us also encouraged the cheap publication of the standard - in lieu
: of a free standard. I think that US$18 is chep enough for most who needs
: a copy.

Sure. But the current distribution and format is non-optimal. We're
programmers. As with anything else we do, we like to see non-optimal stuff,
especially non-optimal stuff we use a lot, made better. Take a question
like "How can we do an XML version of the C++ standard?" I'm sure there is
enough collective effort out there to do this with no cost to any official
body. But if a number of us get together and do this starting from the PDF
document, we're probably violating the terms of our license with ANSI
before we even distribute the first copy. And the whole notion that not
only will the fruit of such work not be distributable, but ANSI probably
wouldn't even offer to sell it makes it seem like a waste of time.

I find it amusing that you suggest a book. Sure I buy books for some
things. But it is not particularly useful compared to an electronic copy
for most of the stuff I use language standards for. (And yes, I end up
looking at the standard a few times a week and I'm not a compiler or
runtime implementor.)

-Z-


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

AllanW@my-dejanews.com writes:

>In article <7bh3go$pir$1@engnews1.eng.sun.com>,
>  stephen.clamage@sun.com (Steve Clamage) wrote:
>> That said, there are some line irregularities, and yes, the
>> typesetting software is not state-of-the-art. (It's troff, used
>> to generate postscript, which is then converted to PDF via
>> the Acrobat Distill program.)

>Any chance that those of us who purchased the PDF version could
>download the troff and/or postscript versions?

Sorry, no. Under ANSI and ISO rules, those docs are not allowed
to be distributed outside the committee.

Personally, I'd like to see an alternative mechanism for
distributing the standards documents, one that would allow any
useful format to be produced.

For example, ISO could license commerical companies to replicate
and distribute standards. I'm sure that major C++ vendors would
like to have the right to include the standard as part of
their C++ distribution.

A while back, some people tried to find out from ISO what it
would take to get such a license. They were unable to get any
sort of an answer, not even whether ISO would consider the
possibility. Maybe they have since received a reply, but if
so, I haven't heard about it.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7baoaf$f0b$1@remarQ.com>, Joe Buck <jbuck@synopsys.com>
writes
>Yes, the standards organizations need a funding mechanism.
>I'm not convinced that the current method is the only option.

So propose an alternative.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

It sounds like ANSI/ISO could use some BPR to get in touch with today's
technology and at least cost effective communications tech.

My question is though: Can the standards participants be persuaded
(required?) to produce a _reference_implementation_?  Has that been
talked about at all?  I believe it would lead to better (a practicianer's
std vs. a computer scientist's std, if you will) standards.

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 3 Mar 1999 22:05:29 GMT, Steve Clamage <Stephen.Clamage@Eng.Sun.COM> wrote:

>For example, ISO could license commerical companies to replicate
>and distribute standards. I'm sure that major C++ vendors would
>like to have the right to include the standard as part of
>their C++ distribution.

Then what would egcs do?

--
----------------------------------
Siemel B. Naran (sbnaran@uiuc.edu)
----------------------------------
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

sbnaran@bardeen.ceg.uiuc.edu (Siemel Naran) writes:

>On 3 Mar 1999 22:05:29 GMT, Steve Clamage <Stephen.Clamage@Eng.Sun.COM> wrote:

>>For example, ISO could license commerical companies to replicate
>>and distribute standards. I'm sure that major C++ vendors would
>>like to have the right to include the standard as part of
>>their C++ distribution.

>Then what would egcs do?

What do you mean? Anyone who wanted to could negotiate the cost
of licensing the right to distribute the standard. In the case
of a product without an owner, you'd be no worse off than you
are now. Individuals who wanted a copy of the standard could
buy one from whoever was selling them.

I didn't mean that the only way you could get a copy of the
standard would be to buy a C++ compiler. I just meant that, for
example, a compiler vendor who wanted to include a copy of the
standard in the distribution would be able so do so.

It used to be common for vendors to distribute a copy of a C++
textbook with the compiler. That didn't mean you couldn't
buy a copy of that texbook from the publisher.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:

>My question is though: Can the standards participants be persuaded
>(required?) to produce a _reference_implementation_?  Has that been
>talked about at all?  I believe it would lead to better (a practicianer's
>std vs. a computer scientist's std, if you will) standards.

Currently, a number of implementors are working at top speed to
produce implementations that are as complete and correct as they
can possibly make them.

So far, no one has succeeded in making one known to be complete
and correct. Even for just the parts of the standard that are
implemented, after the implementation passes all known test
suites, bug reports still come in from users in the field.

Where would the reference implementation come from, and how
would we know it is correct?

IMHO, a reference implementation of a non-trivial language is
possible only when it itself IS the standard. Otherwise, what
do you do when the reference implementation and the published
standard seem to disagree?

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <MPG.1147727523fcd224989b4a@netnews.worldnet.att.net>, Tony
<Tony@ask.me> writes
>My question is though: Can the standards participants be persuaded
>(required?) to produce a _reference_implementation_?  Has that been
>talked about at all?  I believe it would lead to better (a practicianer's
>std vs. a computer scientist's std, if you will) standards.

If I understand what you mean (a model compiler such as that which
exists for C -- as a commercial product) do you even begin to understand
how much work that involves?  BTW standards participants do so on a
voluntary basis  and I can think of few quicker ways of reducing the
number of participants:)



Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Steve Clamage wrote:
>
> Tony@ask.me (Tony) writes:
>
> >My question is though: Can the standards participants be persuaded
> >(required?) to produce a _reference_implementation_?  Has that been
> >talked about at all?  I believe it would lead to better (a practicianer's
> >std vs. a computer scientist's std, if you will) standards.
>
> Currently, a number of implementors are working at top speed to
> produce implementations that are as complete and correct as they
> can possibly make them.
>
> So far, no one has succeeded in making one known to be complete
> and correct. Even for just the parts of the standard that are
> implemented, after the implementation passes all known test
> suites, bug reports still come in from users in the field.
>
> Where would the reference implementation come from, and how
> would we know it is correct?

A reference implementation could do something that a commercial
implementation could not: be horribly inefficient. You could take
advantage of that freedom, to produce a reference implementation much
more quickly than any usuable implementation, solely for the purpose of
proving that the standard is in fact implementable.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Joe Buck wrote:
>
> The issue is not the dollar amount, but the fact that there is a
> dollar amount and that is forbidden to share the document with
> anyone who has not paid.

Or to share it with yourself for that matter. Has anyone actually
read the license at webstore.ansi.org? As I read it, I can't move
my copy of the standard from one machine to another if I still
have access to the original machine. I can't keep a copy at work
and access it by dialing up. Indeed, I can't keep a copy at work
at all. I can't store it in my account on a server and look at it
from a random workstation. I can't put it on a CD and carry that
around with me.

https://webstore.ansi.org/ansidocstore/license/endlicense.asp

"[1]b. You may install one copy of the Product on, and permit
access to it by, a single computer owned, leased or otherwise
controlled by you. In the event that computer becomes
dysfunctional, such that you are unable to access the Product,
you may transfer the Product to another computer, provided that
the Product is removed from the computer from which it is
transferred and the use of the Product on the replacement
computer otherwise complies with the terms of this Agreement.
Neither concurrent use on two or more computers nor use in a
local area network or other network is permitted."

John L

Andrew Koenig wrote:
>
> In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
> Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:
>
> > Here's a suggestion for the next C++ standard: DON'T develop it
> > through the ISO process.  Instead, establish a separate non-profit
> > volunteer "free standards" organization and develop it through that.
> > Release the standard freely and distribute it on the internet.
>
> We thought of that, actually, but it doesn't work--for entirely
> political reasons:  Any collection of vendors that got together
> to agree on a standard for something as important as a major
> programming language would instantly be sued for antitrust!

Even if _everyone_ is allowed to join? And if the resulting
standard is guaranteed to be usable by anyone without any
restrictions, and without having to pay a fee to anyone for
that?
What if the current results are public at any point of the process?

[...]
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Bjarne Stroustrup wrote:

>
> Given that, I think the US$18 that ANSI charges for the pdf version of
> the C++ standard is reasonably cheap. If that's too much, get the free
> slightly-out-of date draft that the C++ standards committee issued for
> public comment (you can find links to both on my home pages).

US$18 is indeed quite reasonable. IEEE is charging way overboard in compa=
rison
to ANSI, check this out:

> 1003.0-1995 IEEE Guide to the POSIX=AE Open System Environment (OSE) (I=
dentical to
> ISO/IEC TR 14252)
> Print: 288 pages [1-55937-531-0] [SH94295-NYF] $98.00 * IEEE Mbr: $78.0=
0
> PDF: [0-7381-0630-5] [SS94295-NYF] $147.00 * IEEE Mbr: $118.00
>
> 9945-1 : 1996 (ISO/IEC) [IEEE/ANSI Std 1003.1, 1996 Edition] Informatio=
n
> Technology--Portable Operating System Interface (POSIX=AE)--Part 1: Sys=
tem Application:
> Program Interface (API) [C Language]
> This edition incorporates extensions for realtime applications (1003.1b=
-1993, 1003.1i-1995)
> and threads (1003.1c-1995).
> Print: 784 pages [1-55937-573-6] [SH94352-NYF] $143.00 * IEEE Mbr: $114=
.00
> PDF: [0-7381-0656-9] [SS94352-NYF] $215.00 * IEEE Mbr: $172.00
>

ANSI is quite reasonable in this regard.


Jonah Tsai
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <F80zoy.Hny@research.att.com>, ark@research.att.com says...

[ ... ]

> We thought of that, actually, but it doesn't work--for entirely
> political reasons:  Any collection of vendors that got together
> to agree on a standard for something as important as a major
> programming language would instantly be sued for antitrust!
> I am told that there is a special dispensation in Federal law
> for ``official'' standards bodies.

This begs the question "How official is official?"  I wonder if
forming something as, say, a subcommittee of the IETF would be
considered legal or not.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

> In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
> Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:
> > Here's a suggestion for the next C++ standard: DON'T develop it
> > through the ISO process.  Instead, establish a separate non-profit
> > volunteer "free standards" organization and develop it through that.
> > Release the standard freely and distribute it on the internet.

He followed it with this important information, which somehow didn't
get quoted properly in follow-up messages:
> > THEN, once the standard is freely available, submit it to ISO via
> > the fast track mechanism (which only allows for a YES/NO vote).
> > I believe that this approach would maximize the benefits to the
> > C++ community.
This quote is important, because without ISO acceptance the new
"standard" would have very little following.

In article <F80zoy.Hny@research.att.com>,
  ark@research.att.com (Andrew Koenig) wrote:
> We thought of that, actually, but it doesn't work--for entirely
> political reasons:  Any collection of vendors that got together
> to agree on a standard for something as important as a major
> programming language would instantly be sued for antitrust!
> I am told that there is a special dispensation in Federal law
> for ``official'' standards bodies.

This can't apply if it's not held in the United States. Besides,
if such a meeting were publicly announced and membership was open
to any interested party, then even if such an antitrust lawsuit
made it to court it would have to fail, wouldn't it?

(Note that I am not a lawyer, so take the above as only the
opinion of an American citizen who naively thinks that U.S. laws
were created to protect innovation, and not to cripple it.)

Or, simply have Fergus Henderson's "separate non-profit volunteer
'free standards' organization" apply for "official standard body"
status.

I think that either of these suggestions makes perfect sense. And
from what I know about politics, making perfect sense dooms it to
failure.

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

>
> Then what would egcs do?
>

I wasn't ware that egcs gave a hoot about the standard at
all.



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

"Al Stevens" <alstevens@midifitz.com> writes:

>Joe Buck wrote in message <7baoaf$f0b$1@remarQ.com>...
>>The issue is not the dollar amount, but the fact that there is a
>>dollar amount and that is forbidden to share the document with
>>anyone who has not paid.
...
>If you were allowed to do that, they would sell about one copy which would
>then be posted for all to retrieve. The product was not financed with public
>funding and its proprietors have no duty to place it in the public domain.
>It was privately funded by the organization and its volunteer members, and
>they have every right to require payment. We are fortunate that they did not
>decide to ask an exhorbitant price.

The trouble is that the interests of the organization and the interests
of its volunteer members do not coincide.  The development of the standard
was primarily funded by the volunteer members, not by ISO.  I think most
of the volunteer members that developed this standard would be happy to
see it freely available.

--
Fergus Henderson <fjh@cs.mu.oz.au>  |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>  |  of excellence is a lethal habit"
PGP: finger fjh@128.250.37.3        |     -- the last words of T. S. Garp.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36DE6305.3EC49DB6@physik.tu-muenchen.de>,
Christopher Eltschka  <celtschk@physik.tu-muenchen.de> wrote:

> > We thought of that, actually, but it doesn't work--for entirely
> > political reasons:  Any collection of vendors that got together
> > to agree on a standard for something as important as a major
> > programming language would instantly be sued for antitrust!

> Even if _everyone_ is allowed to join? And if the resulting
> standard is guaranteed to be usable by anyone without any
> restrictions, and without having to pay a fee to anyone for
> that?

Yes, that is my understanding.

Moreover, even if my understanding were incorrect, it wouldn't help.
For example, the reason that the C++ standard got under way in the
first place was that one compiler vendor had a number of customers
who said that they were not permitted to use C++ because it wasn't
covered by an official standard.
--
    Andrew Koenig
    ark@research.att.com
    http://www.research.att.com/info/ark



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

James Kuyper <kuyper@wizard.net> writes:

>Steve Clamage wrote:
>>
>> Tony@ask.me (Tony) writes:
>>
>> >My question is though: Can the standards participants be persuaded
>> >(required?) to produce a _reference_implementation_?  Has that been
>> >talked about at all?  I believe it would lead to better (a practicianer's
>> >std vs. a computer scientist's std, if you will) standards.
>>
>> Currently, a number of implementors are working at top speed to
>> produce implementations that are as complete and correct as they
>> can possibly make them.
>>
>> So far, no one has succeeded in making one known to be complete
>> and correct. Even for just the parts of the standard that are
>> implemented, after the implementation passes all known test
>> suites, bug reports still come in from users in the field.
>>
>> Where would the reference implementation come from, and how
>> would we know it is correct?

>A reference implementation could do something that a commercial
>implementation could not: be horribly inefficient. You could take
>advantage of that freedom, to produce a reference implementation much
>more quickly than any usuable implementation, solely for the purpose of
>proving that the standard is in fact implementable.

Evidently you have never attempted to implement a C++ compiler or
the C++ standard library.

Efficiency is not the problem.

Correctness is the problem. It is horrendously difficult.

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <F832uJ.EFK@research.att.com>, ark@research.att.com says...
> In article <MPG.1147727523fcd224989b4a@netnews.worldnet.att.net>,
> Tony <Tony@ask.me> wrote:
>
> > My question is though: Can the standards participants be persuaded
>
> Sure thing -- just offer to pay them enough to make it worth their while.
> A few million bucks should do the trick.

That should be easily doable just by using the money that is
currently being used to congregate (someone said 8k per person per
trip!).  Have all interactions occur on the net, dropping expenses to
almost nil.  :)

Tony


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 04 Mar 99 18:22:49 GMT, Francis Glassborow
<francis@robinton.demon.co.uk> enlightened us with:

<snip>
>If I understand what you mean (a model compiler such as that which
>exists for C -- as a commercial product) do you even begin to understand
>how much work that involves?  BTW standards participants do so on a
>voluntary basis  and I can think of few quicker ways of reducing the
>number of participants:)

This is a C++ newsgroup, but most of you people are probably familiar
with what has happened to Fortran over the years. Yes good old
standards meetings, this lead to (I think the germans) requesting
adjustable length strings as part of fortran 95 (or was that 2000?),
anyway only one (I think) vendor spent time (a LOT of time) putting
this into the compiler, everyone made a fuss, the only real standard
is probably fortran 77 (maybe).

Now as an 8 hour/day C++ coder (as I hope you people are), the stl is
great for standardising containers(etc), which means the arguments
over coffee are all about 4,3 or 2 space indents & the odd idiot that
wants 8.
But compliant code?
"If it compiles on the HP, the Alpha, MSVC6, and that linux box then
you can use it".
You see that is why standards organisations exist & people get sent to
attend the meetings, companies would rather spend money sending folks
off for a week to ensure the features that are used/wanted are cast in
stone to save lots and LOTS of money maintaining code futher
downstream.

If a model compiler existed in a free-ish implementation (a.k.a. gcc +
$10 million) the benefits to mankind would probably stretch into the
$billions**. It would also put a few compiler vendors out of business,
but that is probably no bad thing ;)

** Say I develop code that is sold for $300,000 per year. Removal of
headaches would probably amount to 2% of my time = $6000. I'm sure
there must be > 150,000 C++ coders? (Hmmn probably not..)
simon


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Tony@ask.me (Tony) writes:
>> My question is though: Can the standards participants be persuaded
>> (required?) to produce a _reference_implementation_?  Has that been
>> talked about at all?  I believe it would lead to better (a
>> practicianer's std vs. a computer scientist's std, if you will)
>> standards.

Steve Clamage wrote:
> Currently, a number of implementors are working at top speed to
> produce implementations that are as complete and correct as they
> can possibly make them.
>
> So far, no one has succeeded in making one known to be complete
> and correct. Even for just the parts of the standard that are
> implemented, after the implementation passes all known test
> suites, bug reports still come in from users in the field.
>
> Where would the reference implementation come from, and how
> would we know it is correct?
>
> IMHO, a reference implementation of a non-trivial language is
> possible only when it itself IS the standard. Otherwise, what
> do you do when the reference implementation and the published
> standard seem to disagree?

There's also the small problem of time that requiring a reference
implementation would cost.  If the ISO committee had to provide
a working implementation (no matter how inefficient), how much
time would people be willing to wait in order to delay the
acceptance of the standard while the implementation was being
written?  Bear in mind also that the implementation would probably
need to be written by committee (or groups of subcommittees),
simply because of the size and scope of the language and library,
which would add further delays when it came time to integrate the
pieces.

The ISO standard was already delayed about two years (while the
STL was being rolled into it); would anyone have wanted to wait even
longer for final acceptance?

-- David R. Tribble, dtribble@technologist.com --
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <F80zoy.Hny@research.att.com>,
Andrew Koenig <ark@research.att.com> wrote:
>
>We thought of that, actually, but it doesn't work--for entirely
>political reasons:  Any collection of vendors that got together
>to agree on a standard for something as important as a major
>programming language would instantly be sued for antitrust!
>I am told that there is a special dispensation in Federal law
>for ``official'' standards bodies.

Having spent the last 6 years working with an industry association on the
standardization of a protocol where there was no "official" sanction, I
think whoever advised you on that was exagerating.  In order to sue for
antitrust you must be able to show damages.  The antitrust legislation then
allows for multiplied damages.  If there are no restrictions on membership
then nobody can demonstrate damages.  In the case of the industry forum I
was working on some portions of the specification were also published by the
IETF, so gaining that imprimatur, but most of it was just published by the
largest participant in the forum.  All of the specifications resulting from
that forum were put up on the web for free precisely because there was no
need to finance some feather-bedding, pork-barreling, so-called "standards"
organization.

--
Jim Cobban   |  jcobban@nortel.ca                   |  Phone: (613) 763-8013
Nortel Networks (MED)                               |  FAX:   (613) 763-5199
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bks6a$o9u$1@engnews1.eng.sun.com>, stephen.clamage@sun.com
says...
>
> Tony@ask.me (Tony) writes:
>
> >My question is though: Can the standards participants be persuaded
> >(required?) to produce a _reference_implementation_?  Has that been
> >talked about at all?  I believe it would lead to better (a practicianer's
> >std vs. a computer scientist's std, if you will) standards.
>
> Currently, a number of implementors are working at top speed to
> produce implementations that are as complete and correct as they
> can possibly make them.
>
> So far, no one has succeeded in making one known to be complete
> and correct. Even for just the parts of the standard that are
> implemented, after the implementation passes all known test
> suites, bug reports still come in from users in the field.

What do the above 2 paragraphs have to do with the standards _committee_
producing a reference implementation?

> Where would the reference implementation come from, and how
> would we know it is correct?

I thought my question obviously stated that those who are "throwing the
requirements over the wall to engineering/production" would assume the
responsibility/rise to the occasion.  Have you heard of concurrent
engineering?  There is a definite point in time that occurs on a project
after which, if you persist in separating specification from
implementation, much potentially useful information gets lost.  (And
probably a whole host of other unbeneficial effects occur too).

> IMHO, a reference implementation of a non-trivial language is
> possible only when it itself IS the standard.

No, the behavioral spec is still the high level standard, the reference
implementation is the non-value added result that served to concurrently
verify and add quality to the specification as it was being developed.

Now then, in the above scenario, what happens when a value-added
implementor has a question about the spec?  He first goes to the
reference implementation.

Implicit too, in providing a reference implementation, is the fact that
with a reference implementation, less of a specification is required (a
picture is worth a thousand words).  This may be stretching it a bit for
a language spec [product] though.

> Otherwise, what
> do you do when the reference implementation and the published
> standard seem to disagree?

?  Based upon the above dialog, why would they disagree?  By definition,
they couldn't.  The project is done when they finally agree.  The
implementation is defined by the specification and the specification
[project] is influenced, even greatly perhaps, by the implementation
[project].  Analogies bridging academia and industry come to mind.

Tony
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bl907$5o6$1@news-1.news.gte.net>, John Lacey
<johnl@vizdom.com> writes
>"[1]b. You may install one copy of the Product on, and permit
>access to it by, a single computer owned, leased or otherwise
>controlled by you. In the event that computer becomes
>dysfunctional, such that you are unable to access the Product,
>you may transfer the Product to another computer, provided that
>the Product is removed from the computer from which it is
>transferred and the use of the Product on the replacement
>computer otherwise complies with the terms of this Agreement.
>Neither concurrent use on two or more computers nor use in a
>local area network or other network is permitted."

Interesting, exactly how am I supposed to transfer it from a
dysfunctional computer?

I (like quite a few these days) am the sole user of a small LAN.  I
think this is another example of copyright claims being out of phase
with reality (check the copyright restrictions on half a dozen books on
programming and you will find several that do not allow you to transfer
the source code from the book to a computer).


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Christopher Eltschka wrote:
>
> Andrew Koenig wrote:
> >
> > In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
> > Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:
> >
> > > Here's a suggestion for the next C++ standard: DON'T develop it
> > > through the ISO process.  Instead, establish a separate non-profit
> > > volunteer "free standards" organization and develop it through that.
> > > Release the standard freely and distribute it on the internet.
> >
> > We thought of that, actually, but it doesn't work--for entirely
> > political reasons:  Any collection of vendors that got together
> > to agree on a standard for something as important as a major
> > programming language would instantly be sued for antitrust!
>
> Even if _everyone_ is allowed to join? And if the resulting
> standard is guaranteed to be usable by anyone without any
> restrictions, and without having to pay a fee to anyone for
> that?
> What if the current results are public at any point of the process?

In article <MPG.1148da60366d462f989b53@netnews.worldnet.att.net>, Tony
<Tony@ask.me> writes
>That should be easily doable just by using the money that is
>currently being used to congregate (someone said 8k per person per
>trip!).  Have all interactions occur on the net, dropping expenses to
>almost nil.  :)

Sorry (yes I see the smiley) but that really does not work people still
need to meet (and much of the graft is already done over the Internet).
Anyway how does that cut the cost of time.  The real cost is not the
airfares and hotel bills but in the work time spent away from the
employers product.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Andrew Koenig wrote:
....
> For example, the reason that the C++ standard got under way in the
> first place was that one compiler vendor had a number of customers
> who said that they were not permitted to use C++ because it wasn't
> covered by an official standard.

That's the reason why my employer's not been allowed to use it in
delivered software. I'm not sure how long it will take for that to be
changed, now that there is a C++ standard.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Steve Clamage wrote:
>
> James Kuyper <kuyper@wizard.net> writes:
....
> >A reference implementation could do something that a commercial
> >implementation could not: be horribly inefficient. You could take
> >advantage of that freedom, to produce a reference implementation much
> >more quickly than any usuable implementation, solely for the purpose of
> >proving that the standard is in fact implementable.
>
> Evidently you have never attempted to implement a C++ compiler or
> the C++ standard library.

Perfectly true. The closest I've come was a lex/yacc-based program which
parsed just barely enough standard C syntax to create a header file with
extern declarations for all globally visible symbols that were defined
in the input file.

> Efficiency is not the problem.

I'm surprised. I've heard  KAI C++, in particular, described as having
"almost supernatural" optimization capabilities. I would have expected
that achieving that level of optimization would require far more work
than merely correctly interpreting the source code.

> Correctness is the problem. It is horrendously difficult.

I'll believe that.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36DF3EB5.3CE14A6D@wizard.net>,
James Kuyper  <kuyper@wizard.net> wrote:
>I think most people reading this newsgroup are likely to agree that
>standardization is a good idea, but there needs to be a barrier
>preventing it from being misused. For example, consider what a
>Microsoft-dominated committee would have done. In the opposite
>direction, consider how Microsoft feels about Sun's sponsorship of the
>work on a Java standard.

ANSI and ISO rules are supposed to prevent things like a
"Microsoft-dominated committee".  In ANSI committees, the rule is "one
company, one vote"; in ISO it's "one country, one vote".

--
Barry Margolin, barmar@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Francis Glassborow <francis@robinton.demon.co.uk> writes:

> ...people still
>need to meet (and much of the graft is already done over the Internet).

I sincerely hope that Francis meant "draft".

I am NOT a crook! :-)

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 02 Mar 99 06:19:27 GMT, clamage@Eng.Sun.COM (Steve Clamage) wrote:

>Other people have complained about the font, but so far no one
>has said what specifically they object to.

Apparently, the typesetting software used to layout the document
is not very smart. It has serious problems with character spacing,
kerning, etc. In addition, because C++ names are typeset in Courier,
and the paragraphs were justified, the lines are very uneven in some
places, which reduces readability greatly.

---------------------------------------------
Ziv Caspi
zivca@netvision.net.il
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

zivc@peach-networks.com (Ziv Caspi) writes:

>On 02 Mar 99 06:19:27 GMT, clamage@Eng.Sun.COM (Steve Clamage) wrote:

>>Other people have complained about the font, but so far no one
>>has said what specifically they object to.

>Apparently, the typesetting software used to layout the document
>is not very smart. It has serious problems with character spacing,
>kerning, etc. In addition, because C++ names are typeset in Courier,
>and the paragraphs were justified, the lines are very uneven in some
>places, which reduces readability greatly.

Assuming we are still talking about the PDF and not the printed
document, such details are highly dependent on the computer on
which Acrobat runs, the display driver, and to some degree on
which version of Acrobat you are using.

During development of the standard, we saw cases where a given
PDF version of a draft looked different on different computers.
In a particularly bizarre case, we saw some pages display upside
down (!) on one computer, although the same file displayed
correctly on other computers.

That said, there are some line irregularities, and yes, the
typesetting software is not state-of-the-art. (It's troff, used
to generate postscript, which is then converted to PDF via
the Acrobat Distill program.)

--
Steve Clamage, stephen.clamage@sun.com


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bh3go$pir$1@engnews1.eng.sun.com>,
  stephen.clamage@sun.com (Steve Clamage) wrote:
> That said, there are some line irregularities, and yes, the
> typesetting software is not state-of-the-art. (It's troff, used
> to generate postscript, which is then converted to PDF via
> the Acrobat Distill program.)

Any chance that those of us who purchased the PDF version could
download the troff and/or postscript versions?

----
AllanW@my-dejanews.com is a "Spam Magnet" -- never read.
Please reply in USENET only, sorry.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:

> Here's a suggestion for the next C++ standard: DON'T develop it
> through the ISO process.  Instead, establish a separate non-profit
> volunteer "free standards" organization and develop it through that.
> Release the standard freely and distribute it on the internet.

We thought of that, actually, but it doesn't work--for entirely
political reasons:  Any collection of vendors that got together
to agree on a standard for something as important as a major
programming language would instantly be sued for antitrust!
I am told that there is a special dispensation in Federal law
for ``official'' standards bodies.
--
    Andrew Koenig
    ark@research.att.com
    http://www.research.att.com/info/ark



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 26 Feb 99 19:02:09 GMT, bs@research.att.com (Bjarne Stroustrup)
wrote:
>Given that, I think the US$18 that ANSI charges for the pdf version of
>the C++ standard is reasonably cheap.

I believe that it has also set a new record as the lowest price ever
for a copy of an ANSI standard -- as close to "free" as any ANSI
standard has ever been. It is also, I believe, the first ANSI standard
to be made officially available in electronic form.

This was quite a double-barrelled accomplishment. It was achieved
through the hard work and lobbying of many dedicated people ISO and
ANSI people, some of whom participate in this newsgroup.

Herb


---
Herb Sutter (mailto:hsutter@peerdirect.com)

PeerDirect Inc.     2695 North Sheridan Way, Suite 150
www.peerdirect.com  Mississauga Ontario Canada L5K 2N6


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In this next-to-last year of the millennium, hsutter@peerdirect.com (Herb
Sutter) wrote in article <36d84c43.15501981@nr1.toronto.istar.net> in
comp.std.c++:
>On 26 Feb 99 19:02:09 GMT, bs@research.att.com (Bjarne Stroustrup)
>wrote:
>>Given that, I think the US$18 that ANSI charges for the pdf version of
>>the C++ standard is reasonably cheap.
>
>I believe that it has also set a new record as the lowest price ever
>for a copy of an ANSI standard -- as close to "free" as any ANSI
>standard has ever been. It is also, I believe, the first ANSI standard
>to be made officially available in electronic form.
>
>This was quite a double-barrelled accomplishment. It was achieved
>through the hard work and lobbying of many dedicated people ISO and
>ANSI people, some of whom participate in this newsgroup.

Well said!

The standard is available for much less than the cost of most printed
computer books. I am sure that I'm far from the only person who is very
grateful.

(To moderator: sorry for the lengthy quoting. I just couldn't bear to
trim any of the above.)

--
Stan Brown, Oak Road Systems, Cleveland, Ohio, USA
                                    http://www.mindspring.com/~brahms/
My reply address is correct as is. The courtesy of providing a correct
reply address is more important to me than time spent deleting spam.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

>>Why isn't the C++ standart not free? That kind of information should be
>>public to encourage better, standart complaint code!!

Francis Glassborow <francisG@robinton.demon.co.uk> writes:
>And exactly who do you propose pay for the vast quantity of work that
>goes into Standards development?

You appear to be under the mistaken belief that the money collected for
copies of the standards documents pays for the work that created the
standard.  It does not.  Almost all the expenses of standards creation
are borne by the employers of the volunteers that developed the standard.
The standards organizations have costs, yes, but we are long overdue
for re-thinking about how the standards organizations are funded.

>I think that $18 is a pretty small price for electronic access to a
>document that has literally cost millions to produce.

Since that $18 doesn't go to the people who spent the lion's share of
the money, it doesn't matter.  Do you think that Andy Koenig or Steve
Clamage is going to see any of that money, or that it will be used to
compensate their employers for the time they devoted to its production
instead of building something their companies could sell?  Of course not.

We could figure out how to raise the money some other way.  Companies
pay the expenses to get the volunteers to meetings because it is valuable
to have a say in what the standard will be.  Other organizations besides
companies (governments, nonprofits, universities) also have an interest.
All we need to do is to compute how much to raise the cost of
participation by to make up the difference so that we can afford to
distribute the standards on the Internet for zero cost.  The copyright
would permit republishing and quoting of passages but not modification;
the fear expressed by some that people might publish false standards
is solvable by digital signatures.  Of course people who want paper
copies would still pay for the service of producing a copy.



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

> I believe that it has also set a new record as the lowest price ever
> for a copy of an ANSI standard -- as close to "free" as any ANSI
> standard has ever been. It is also, I believe, the first ANSI standard
> to be made officially available in electronic form.
>

Actually, neither of those statements are true.  ANSI standards
have been made available in electronic form before and they're
are some that cost nothing.

However, in the realm of useful computer specs, you are right,
$18 is a bargain.

My only real gripe is that they disabled cutting and pasting and
the conversion from TROFF to PDF was not without problems.
It's no where near as nice as the HTML'ized drafts where the
links actually work.



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 26 Feb 99 19:02:09 GMT, bs@research.att.com (Bjarne Stroustrup)
>wrote:
>>Given that, I think the US$18 that ANSI charges for the pdf version of
>>the C++ standard is reasonably cheap.

hsutter@peerdirect.com (Herb Sutter) writes:
>I believe that it has also set a new record as the lowest price ever
>for a copy of an ANSI standard -- as close to "free" as any ANSI
>standard has ever been.

The issue is not the dollar amount, but the fact that there is a
dollar amount and that is forbidden to share the document with
anyone who has not paid.

If the fee were $0.01, that still wouldn't solve the problem.
The problem is not that we must pay for a PDF copy of the standard.
It is that the standard cannot be placed on the net, with hyperlinks,
commentary, and the like.  You cannot instantly respond to a student's
question about some tricky C++ issue by just giving them a URL,
which you know to be valid even if you are in Japan and she is in
Slovenia (unless the language has not changed from CD2).

Yes, the standards organizations need a funding mechanism.
I'm not convinced that the current method is the only option.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Joe Buck wrote in message <7baoaf$f0b$1@remarQ.com>...
>The issue is not the dollar amount, but the fact that there is a
>dollar amount and that is forbidden to share the document with
>anyone who has not paid.
>
>If the fee were $0.01, that still wouldn't solve the problem.
>The problem is not that we must pay for a PDF copy of the standard.
>It is that the standard cannot be placed on the net, with hyperlinks,
>commentary, and the like.  You cannot instantly respond to a student's
>question about some tricky C++ issue by just giving them a URL,
>which you know to be valid even if you are in Japan and she is in
>Slovenia (unless the language has not changed from CD2).

If you were allowed to do that, they would sell about one copy which would
then be posted for all to retrieve. The product was not financed with public
funding and its proprietors have no duty to place it in the public domain.
It was privately funded by the organization and its volunteer members, and
they have every right to require payment. We are fortunate that they did not
decide to ask an exhorbitant price.

>Yes, the standards organizations need a funding mechanism.
>I'm not convinced that the current method is the only option.

Short of a bake sale, I don't know what another option could be.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Hi,
Joe Buck (jbuck@synopsys.com) wrote:
: Francis Glassborow <francisG@robinton.demon.co.uk> writes:
: >And exactly who do you propose pay for the vast quantity of work that
: >goes into Standards development?

: You appear to be under the mistaken belief that the money collected for
: copies of the standards documents pays for the work that created the
: standard.  It does not.

Since Francis is a member of the C++ Standardization Committee he knows
this very well! All he is saying is that other people, that is those
involved in the creation of the standard, are already paying quite a
lot for the standard such that $18 is definitely an acceptable price.

: We could figure out how to raise the money some other way.  Companies
: pay the expenses to get the volunteers to meetings because it is valuable
: to have a say in what the standard will be.  Other organizations besides
: companies (governments, nonprofits, universities) also have an interest.
: All we need to do is to compute how much to raise the cost of
: participation by to make up the difference so that we can afford to
: distribute the standards on the Internet for zero cost.

A rough calculation results in something like costs of $8000 per
meeting (at least $5000 lost money due to the absence of the person,
$2000 travel expenses, $800 for ANSI membership, and the rest is
definitely made up by the necessary administration). You want to charge
those people more? That's great: It would give me hard time to convince
my boss to pay another $8000 just because others want to get the
standard for free! My whole interest in the standard is to make C++ a
reasonable language. The only interest for my company in C++ is my
knowledge which can be sold.

Here is an alternative approach: Write up a document which translates
the standard into words readable by "normal" programmers, potentially
adding a rationale, and publish it for free. I'm pretty sure that may
people in the committee are willing to review it for correctness. This
document can then be used for reference. I see absolutely no reason why
all the costs have to be paid by committee members!
--
<mailto:dietmar.kuehl@claas-solutions.de>
<http://www.informatik.uni-konstanz.de/~kuehl/>
I am a realistic optimist - that's why I appear to be slightly pessimistic
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Joe Buck (jbuck@synopsys.com) wrote:
[...]

Joe makes the point that free means much more than "I don't have to pay for
it." The $18 I paid for my PDF copy of the standard was money well spent. I
have no complaint with that. I do have a complaint with the experience
given by the PDF document. As with virtually anything PDF based, it sucks.
You'd think by now PDF documents would reliably match page number
references in the text to page number references in the user interface of
the document viewer. (Other document viewers manage to do this with complex
real world documents...)

If the standard were "free" I'm sure there would already be a number of Web
sites available with annotated versions of the standard, better navigation,
code examples, references to what real compilers do in undefined
situations, debugging advice, etc. FAQ's would have direct links to the
standard. I could answer emails with a URL instead of hand wavingly saying
"I read my personal copy of the standard and it says the code is broken."
Hell, ISO and ANSI might make more money putting the standard up on a Web site
and selling advertising. (No, I don't have an analysis to back that up, but
I don't know how many copies of the standard they've sold either.)

So take this as a request for the standards organization to see this
document as a living thing that must move forward. I'd happily pay another
$18 bucks for a high quality HTML or XML version of the document. My
impression of how standards organizations work leads me to believe we won't
see any such changes until the next revision of the standard itself.

Having only one source for the document which is the C++ standard isn't
very different from having only one source for your compiler or OS
software. The end user experience ultimately suffers, at least for many end
users.

-Z-
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

kuehl@horn.fmi.uni-konstanz.de (Dietmar Kuehl) writes:

>Joe Buck (jbuck@synopsys.com) wrote:
>: Francis Glassborow <francisG@robinton.demon.co.uk> writes:
>: >And exactly who do you propose pay for the vast quantity of work that
>: >goes into Standards development?
>
>: You appear to be under the mistaken belief that the money collected for
>: copies of the standards documents pays for the work that created the
>: standard.  It does not.
>
>Since Francis is a member of the C++ Standardization Committee he knows
>this very well! All he is saying is that other people, that is those
>involved in the creation of the standard, are already paying quite a
>lot for the standard such that $18 is definitely an acceptable price.

I don't think this follows.  $18 is not a large price, but any price
at all leads to a major reduction in the benefit that we all get from
having a standard.  The volunteers are donating their time and money
to give the community a standard.  If the standard were freely available,
that would maximize the benefits that we can achieve from their efforts.
Making the standard freely available is thus a way of maximimizing
the return on the volunteer's investment.

>: We could figure out how to raise the money some other way.  Companies
>: pay the expenses to get the volunteers to meetings because it is valuable
>: to have a say in what the standard will be.  Other organizations besides
>: companies (governments, nonprofits, universities) also have an interest.
>: All we need to do is to compute how much to raise the cost of
>: participation by to make up the difference so that we can afford to
>: distribute the standards on the Internet for zero cost.
>
>A rough calculation results in something like costs of $8000 per
>meeting (at least $5000 lost money due to the absence of the person,
>$2000 travel expenses, $800 for ANSI membership, and the rest is
>definitely made up by the necessary administration). You want to charge
>those people more?

Yes.  But only a tiny tiny fraction more.  All that is required is a little
time and bandwidth for the WWW site.  These people (I am one of them)
would I am sure be happy to volunteer the additional time, computer
resources, and bandwidth.  For example, Bjarne already has a link to the
last public comment version of the standard on his home page; I'm sure
he and many others would be happy to put the standard itself up on their
WWW pages if it were freely available.

The cost of developing the standard must have been roughly about
$5000/person/meeting * about 100 people/meeting * 3 meetings/year *
about 5 years = about 5 or 10 million dollars.  The additional cost
to make it freely available on the WWW would be a miniscule fraction
of that.

>That's great: It would give me hard time to convince
>my boss to pay another $8000 just because others want to get the
>standard for free!

The additional cost to you (and your boss) would not be $8000.
It would in fact be precisely zero, unless you volunteered to help
distribute the standard.  I don't think there would be any shortage
of such volunteers.

Here's a suggestion for the next C++ standard: DON'T develop it
through the ISO process.  Instead, establish a separate non-profit
volunteer "free standards" organization and develop it through that.
Release the standard freely and distribute it on the internet.
Fund the bandwidth and computer resources needed for the distribution
using advertising, if necessary.  THEN, once the standard is freely
available, submit it to ISO via the fast track mechanism (which only
allows for a YES/NO vote).  I believe that this approach would
maximize the benefits to the C++ community.

--
Fergus Henderson <fjh@cs.mu.oz.au>  |  "Binaries may die
WWW: <http://www.cs.mu.oz.au/~fjh>  |   but source code lives forever"
PGP: finger fjh@128.250.37.3        |     -- leaked Microsoft memo.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

On 28 Feb 1999 21:15:49 GMT, Ron Natalie <ron@sensor.com> wrote:

>My only real gripe is that they disabled cutting and pasting and
>the conversion from TROFF to PDF was not without problems.
>It's no where near as nice as the HTML'ized drafts where the
>links actually work.

The PDF version is also slow for scrolling, and the font hurts my
eyes.

--
----------------------------------
Siemel B. Naran (sbnaran@uiuc.edu)
----------------------------------


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7baoaf$f0b$1@remarQ.com>, Joe Buck <jbuck@synopsys.com> wrote:
>It is that the standard cannot be placed on the net, with hyperlinks,
>commentary, and the like.  You cannot instantly respond to a student's
>question about some tricky C++ issue by just giving them a URL,
>which you know to be valid even if you are in Japan and she is in
>Slovenia (unless the language has not changed from CD2).

Note that Kent Pitman, the editor of the ANSI Common Lisp standard, managed
to get ANSI to allow him to place an *unofficial* version of the standard
on the net, complete with hyperlinks, and even allow free copying.  It's
called the Common Lisp HyperSpec.  There's a disclaimer that says that it's
not considered legally equivalent to the ANSI spec, but for most users it's
adequate (it was derived from the same source code as was used for the
standard), and even has additional features not in the ANSI spec (links to
some committee documents, more indexes).

Perhaps you folks should investigate something like this.

--
Barry Margolin, barmar@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36d84c43.15501981@nr1.toronto.istar.net>,
hsutter@peerdirect.com says...

[ $18US PDF C++ STD ]

> I believe that it has also set a new record as the lowest price ever
> for a copy of an ANSI standard -- as close to "free" as any ANSI
> standard has ever been.

Nope -- a quick glance at the ANSI catalog shows at least a couple for
$14US.  However, I strongly suspect these are MUCH smaller standards.

> It is also, I believe, the first ANSI standard
> to be made officially available in electronic form.

This depends a bit on your viewpoint.  The text of the Ada standard is
freely available on-line.  At least as I understand things, this isn't
quite exactly the standard itself though -- it's really a base
document that ANSI happened to turn into a standard without doing an
editing.  If you buy a copy of the Ada standard from ANSI, it's
$210US.  I suppose there are situations where one needs a real bona
fide copy of the standard, and you'd spend that kind of money, but for
most people it would be silly.

> This was quite a double-barrelled accomplishment. It was achieved
> through the hard work and lobbying of many dedicated people ISO and
> ANSI people, some of whom participate in this newsgroup.

It's quite an accomplishment, and many of us appreciate it a great
deal, regardless of the facts above.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:
>Here's a suggestion for the next C++ standard: DON'T develop it
>through the ISO process.  Instead, establish a separate non-profit
>volunteer "free standards" organization and develop it through that.
>Release the standard freely and distribute it on the internet.

Don't forget the step of convincing numerous governments around the world
to approve this as a valid standardization body.  Many governments mandate
that ISO standards must be used when they apply.

--
Barry Margolin, barmar@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Barry Margolin <barmar@bbnplanet.com> writes:

>In article <7bdnlf$640$1@mulga.cs.mu.OZ.AU>,
>Fergus Henderson <fjh@cs.mu.OZ.AU> wrote:
>>Here's a suggestion for the next C++ standard: DON'T develop it
>>through the ISO process.  Instead, establish a separate non-profit
>>volunteer "free standards" organization and develop it through that.
>>Release the standard freely and distribute it on the internet.
>
>Don't forget the step of convincing numerous governments around the world
>to approve this as a valid standardization body.  Many governments mandate
>that ISO standards must be used when they apply.

You have quoted me rather selectively.  I didn't forget that step --
it's not necessary.  You don't need to convince all those governments,
all you need to do is to convince ISO to accept the standard developed
by the new organization.  As I said: "THEN, once the standard is freely
available, submit it to ISO via the fast track mechanism".

--
Fergus Henderson <fjh@cs.mu.oz.au>  |  "Binaries may die
WWW: <http://www.cs.mu.oz.au/~fjh>  |   but source code lives forever"
PGP: finger fjh@128.250.37.3        |     -- leaked Microsoft memo.
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

sbnaran@fermi.ceg.uiuc.edu (Siemel Naran) writes:

>The PDF version is also slow for scrolling, and the font hurts my
>eyes.

Other people have complained about the font, but so far no one
has said what specifically they object to.

The PDF fonts are the fonts in the printed documents -- primarily
Times Roman and Courier. Are you (those who have complained)
objecting to those fonts, or to the way they are rendered by Acrobat?

By default, the Acrobat reader displays text "smoothed". I find the
result not very readable, since it reduces the contrast. If you turn
off smoothing (File/Preferences/General), you get more contrast,
and faster display as well.

--
Steve Clamage, stephen.clamage@sun.com
---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

Why isn't the C++ standart not free? That kind of information should be
public to encourage better, standart complaint code!!
Joao



[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36d5b13c.0@news.ip.pt>,
Joao Sampaio <joao_sampaio@spamhate.bigfoot.com> wrote:
>Why isn't the C++ standart not free? That kind of information should be
>public to encourage better, standart complaint code!!

Because ANSI is a company that gets its revenues from publishing standards.

Customers and industry should encourage standard-compliant code.  ANSI just
facilitates the production and publication of the standards.

--
Barry Margolin, barmar@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]

In article <36d5b13c.0@news.ip.pt>, Joao Sampaio <joao_sampaio@spamhate.
bigfoot.com> writes
>
>Why isn't the C++ standart not free? That kind of information should be
>public to encourage better, standart complaint code!!
>Joao

And exactly who do you propose pay for the vast quantity of work that
goes into Standards development?  As it is many companies and
individuals donate many thousands of dollars each.  For example the
annual cost of travel and accommodation for an individual attending WG21
& J16 meetings is of the order of $5000 and that does not include the
cost in lost work time.

I think that $18 is a pretty small price for electronic access to a
document that has literally cost millions to produce.


Francis Glassborow      Chair of Association of C & C++ Users
64 Southfield Rd
Oxford OX4 1PA          +44(0)1865 246490
All opinions are mine and do not represent those of any organisation


[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://reality.sgi.com/austern_mti/std-c++/faq.html              ]