Standards Update, X3J16: C++
Jeffrey S. Haemer
jsh at usenix.org
Thu Oct 4 00:58:36 AEST 1990
Submitted-by: jsh at usenix.org (Jeffrey S. Haemer)
An Update on UNIX1-Related Standards Activities
October 3, 1990
USENIX Standards Watchdog Committee
Jeffrey S. Haemer <jsh at usenix.org>, Report Editor
X3J16: C++
Mike Vilot <mjv at objects.mv.com> reports on the July meeting in
Seattle, Washington:
Standard C++?
The C++ programming language has been gaining popularity at a
remarkable rate (an informal estimate is that the C++ population
doubles every nine months). One reaction to the growing popularity
has been a call to stabilize the language's definition, and achieve
some consistency across implementations. C++ is popular enough that
larger corporations are considering adopting it as an officially
endorsed development language -- but some cannot make such a move
unless the language is defined by a standard.
For these and other reasons, the ANSI secretariat agreed to establish
the X3J16 committee to formulate a standard for C++. Dmitry Lenkov,
of HP, made the official proposal, and serves as chairman of the
committee. To date, X3J16 has met three times: an organizational
meeting last December, the first technical meeting in March to get
organized, and a meeting in July to really get started.
The December meeting, in Washington D.C., was purely administrative:
over 50 attendees received lectures and tons of paper on X3 rules and
procedures. The highlight of the day was an invited presentation by
Bjarne Stroustrup on ``the spirit of C++.'' The transcript is
available as committee document X3J16/90-0022 or from Greg Comeau at
Comeau Computing, 91-34 120th Street, Richmond Hill, NY 11418, (718)
849-2355.
March meeting
AT&T hosted the meeting in New Jersey. Most of the week was spent on
administrative matters, while the group got organized and accustomed
to The Bureaucratic Way. Since most of the members are engineers, the
highlight of the week was the evening technical sessions on
implementing exception handling for C++. (The week was sort of a
__________
1. UNIXTM is a Registered Trademark of UNIX System Laboratories in
the United States and other countries.
October 3, 1990 Standards Update X3J16: C++
- 2 -
mini-Usenix conference, as most members had gone without a substantial
C++ gathering since the October '88, Denver conference.)
The week's major activities were discussing and preparing a goals
document, describing the committee's activities and priorities.
Goals
Here is a brief outline of the goals document, which is available as
X3J16/90-0023:
1. Base documents: C++ Reference Manual, ANSI C (ANS X3.159-1989),
ISO C when available.
2. Standardize syntax and semantics of the language as a token
sequence without the presence of preprocessing directives.
3. Define and standardize a minimum set of C++ libraries, their
contents, and interfaces.
4. Standardize elements of a C++ environment.
5. Consider proposed major changes: parameterized types and
exceptions.
6. Ensure that the standard is suitable for the international
community.
7. Ensure a very high level of compatibility with ANSI C.
8. Establish coordinating liaisons with X3J11 (ANSI C) and
Numerical C Extensions Group.
9. Produce two deliverables: draft proposed standard and rationale.
10. Priorities:
- clear & unambiguous
- C++ reference manual
- other base documents
- consistency
- user/implementer experience
- portability, efficiency, expressiveness
- ease of implementation (including translation to C)
October 3, 1990 Standards Update X3J16: C++
- 3 -
There was some confusion over the multiple base documents. Most
members had seen the AT&TT C++ version 2.0 reference manual, but in
preparation for standardization, the language and its reference manual
had suffered a number of subsequent, small changes. AT&T made the 2.1
reference manual available to X3J16; it was essentially the text of
the book The Annotated C++ Reference Manual by Margaret Ellis and
Bjarne Stroustrup published by Addison-Wesley (minus the annotations).
My naive suggestion to remove the ANSI C standard as a base document
in favor of a single base provoked the most intense and emotional
discussion of the week. At stake was compatibility between C++ and C.
While most members of X3J16 feel that the existence of a separate
committee implies the standardization of a new language, some former
members of X3J11, which just finished the C standard, want to
eliminate any and all incompatibilities with C. (There was even a
threat to sabotage the C++ standard in balloting if they are not
removed.)
This issue is obviously important and has two sides. Make your
preferences known to the committee. For detailed reference material,
both ``C++: As Close as Possible to C -- But No Closer,'' (Bjarne
Stroustrup and Andy Koenig, The C++ Report, 1(7), 1989) and Chapter 18
of The Annotated C++ Reference Manual document and explain differences
and incompatibilities between the languages as they stand today.
Focusing on a language without preprocessing directives continues the
de-emphasis of the C preprocessor. This is particularly favored by
C++ vendors looking into more powerful development environments.
[Editor: Admittedly, improper preprocessor use can sink us in deep and
dirty bath water, but let's make sure to save the baby. When writing
portable C, I personally find #ifdefs extremely valuable; I suspect
they will remain valuable in C++, and I would hate to see the working
group neglect this valuable porting tool.]
The libraries effort includes asking what to do about the ANSI-C
library, and investigating what can be standardized in a more C++-like
approach.
The environment work addresses the linking and executing of C++ code
with non-C++ code (i.e., linkage and program execution models), rather
than development environment tools.
There are thousands of suggested ``improvements'' proposed as
extensions to C++, but there is consensus on two named in the goals
document: parameterized types and exception handling. Their proposals
are detailed, and both have been implemented (in some form) in a few
C++ implementations.
The emphasis on international concerns reflects the lessons learned
from the difficulties of C standardization. X3J16 has some fences to
October 3, 1990 Standards Update X3J16: C++
- 4 -
mend, particularly in the international community. Rather than
waiting until the last minute to spring a standard on the ISO, the C++
committee is involving itself with the international community right
from the start.
July meeting
Microsoft hosted the second, Seattle meeting. Sub-groups focused on
the key topics listed in the goals statement began work at the March
meeting, and reported their progress in Seattle.
International Concerns
Steve Carter, of Bellcore, presented the major International
Concerns (particularly character sets and formal specification)
and asked the other groups to work on these issues. He also
suggested various sites overseas where future X3J16 meetings
could help cooperation with international standardization
efforts.
Editorial
Jonathan Shopirio, of AT&T, presented the Editorial group's
proposal for organizing and formatting the standard. Jon is also
working on an abstract machine model, and a way to define the
semantics in the standard precisely and consistently.
Formal Syntax
James Roskind, an independent consultant, presented the work of
the Formal Syntax group. He has developed (and published on the
net) a yacc-able grammar for C++, and is concerned about
ambiguities in the the language. Most of the discussion was
spent trying to discover whether C++ can (or should) be made
LALR(1).
Core Language
Andy Koenig, of AT&T, presented the Core Language group's work.
Initially, they identified and classified difficulties in the
working document.
Environment
John Vasta, of HP, presented the work of the Environment group.
A key issue addressed by this group is the interaction of C++
with other programming languages. Among the important topics are
linkage of C++ and non-C++ translation units, especially the
construction and destruction of static C++ objects.
Libraries
I presented the Library group's work. There were many
suggestions, from both inside and outside of the committee.
(Interested outside suggesters were James Coggins, Keith Gorlen,
and Doug Lea, who have each developed large C++ libraries.) A few
people noted similarity with topics covered by other standards
October 3, 1990 Standards Update X3J16: C++
- 5 -
(notably POSIX). Initially, the library group will focus on a
few commonly-used components. Parameterized types and exception
handling will significantly effect the way we design libraries in
C++.
Language Extensions
Bjarne Stroustrup, of AT&T, presented the work of the Extensions
group, which was by far the most active. The technical sessions
presented experience with implementation and use of the template
facility.
The most active and emotional debate of the week was on exception
handling, discussing the proposal outlined by Andy Koenig and
Bjarne Stroustrup in their paper ``Exception Handling for C++''
presented at the Usenix C++ conference in April. Martin
O'Riordan, of Microsoft, and Mike Miller, of Glockenspiel,
presented arguments in favor of extending the current proposal
(which defines termination semantics for exceptions) to include
resumption semantics. Andy and Bjarne explained their reasons
for not including resumption -- the most important was the
complexity and cost of implementation.
To their credit, the group worked hard to find a proposal that
provided both kinds of exceptions with acceptably small
time/space overhead. However, at the end of the week, Bjarne
declared the debate deadlocked, and refused to impose his
proposal while substantial disagreement remained. This is
another topic where you should make your opinions heard.
C Compatibility
Mike Miller presented the work of the C Compatibility group. Tom
Plum, of Plum-Hall, produced a list of every section of the C++
reference manual that was not C. Much of the group's near-term
activity will be devoted to explaining the many items on the
list.
The Seattle meeting produced tangible progress on the language
standard. X3J16 voted to accept the proposed document outline and
format. They also agreed to incorporate the template proposal (the
text from Chapter 14 of The Annotated Reference Manual, minus the
annotations -- it was literally a scissors-and-tape job). We hope C++
vendors will regard templates as now officially in the language, and
provide users an opportunity to work with this feature.
Next events
A few substantial issues lie ahead. The next meeting should see some
resolution on the exception proposal. We should see some progress on
the review of language ambiguities and inconsistencies, and have some
idea of how difficult it will be to ANSIfy the document. We should
also see some specific proposals on library contents. The most
October 3, 1990 Standards Update X3J16: C++
- 6 -
substantial will be a simplified version of iostreams by Jerry Schwarz
(now at Stardent, formerly at AT&T).
Our target date for delivering a draft standard is the end of 1992.
X3J16 meets three times per year. The next three meetings (and their
hosts) will be:
+ November 12-26, Cupertino CA (Hewlett Packard)
+ March 11-15, Nashua NH (Digital)
+ June 17-21, Lund Sweden (Lund Institute of Technology)
Membership on an X3 committee is open to any individual or
organization with expertise and material interest in the topic
addressed by the committee. The cost for membership is $250. Contact
the chair or vice chair for details.
Chair: Dmitry Lenkov
HP California Language Lab
19447 Pruneridge Avenue MS 47 LE
Cupertino, CA 95014
(408)447-5279
FAX (408)447-4924
email dmitry%hpda at hplabs.hp.com
Vice Chair: William M. Miller
Glockenspiel, Ltd
P.O. Box 366
Sudbury, MA 01776-0003
(508)443-5779
email wmmiller at cup.portal.com
October 3, 1990 Standards Update X3J16: C++
Volume-Number: Volume 21, Number 174
More information about the Comp.std.unix
mailing list