SC22/WG21/N1170
J16/98-0027
25 October, 1998
Sean A Corfield
sean@ocsltd.com

Plum called the meeting to order at 18:00PST, 6 October, 1998.

Plum explained that there were three official hosts: Silicon Graphics Inc, Plum Hall Inc and Perennial Inc. He also noted that a lot of the local meeting arrangements had been made by Tom MacDonald of Cray Inc and that Simonsen (Denmark) had providing Internet access and networking courtesy of the Danish UNIX Users Group.

In attendance were:

Plum was acclaimed.

Corfield acted as secretary.

Plum nominated English. Approved by acclamation.

The agenda, N1158 = 98-0015, was adopted by acclamation.

Plum noted that a drafting committee might be needed to draft any New Project proposals that might be agreed during the week. Simonsen volunteered and the Japanese delegation offered to assist him. Kühl also volunteered.

Corfield said he had heard no objections to the minutes of the previous meeting. Plum asked if there were any and no one objected. The minutes, N1154 = 98-0011, were approved by acclamation.

There were no outstanding action items.

Simonsen brought forward a paper on Internationalisation issues and would obtain document numbers from Miller later.

All deferred until joint session.

Plum said we should celebrate the FDIS ballot result. Applause. The ISO C++ Standard document is designated 14882 ("14 8 82"). Simonsen noted that a PDF version is available for download from ANSI’s web site. Plum noted that ANSI's price of $18 for the PDF download is the lowest ever for a standard and that we should encourage people to buy it (if ANSI is their usual Standards Body).

Plum said we will discuss this in full committee but asked if anyone had any particular comments. So far we have two DRs and lots of known issues.

Plum said that, at the SC22 plenary in August, he & Simonsen reviewed the proposal against previous NP proposals and then Kehoe also reviewed it. Plum said "performance" is not just to do with speed, it also covers resource usage etc. O'Riordan said embedded / real-time market is more interested in repeatability than outright performance. Plum feels many of the embedded people who used C are now looking to C++ since C's revision is not focusing on that area. WG21 is therefore the "right place" for this discussion. Japan has already done two years’ work on this subject. Kühl asked whether the work on performance issues would be co-located with the scheduled J16 meetings. Plum said yes.

Simonsen said WG20 - Internationalization - is making a C++ binding and wants to talk to the group. Plum said this is purely a library issue.

Plum agreed to chair the next meeting which will be in Dublin, Ireland, 12-16 April, 1999.

Currently, the only scheduled future meeting is in Kona, Hawaii, 20-26 October 1999. Plum said this would be a five day meeting, co-located with WG14, and covers Wed-Fri (20^th-22^nd) & Mon-Tue (25^th-26^th). Plum suggested an East Coast meeting in April 2000.

Plum moved to adjourn at 19:05PST, 6 October, 1998.

Clamage opened the meeting at 9:06PST on Wednesday 7 October 1998.

Clamage explained voting procedures. Miller circulated the membership list and J16 attendance list and explained the document numbering.

Document number N1159 = 98-0016. Agenda approved by acclamation.

None.

Motion by Miller/Sutter to approve N1154 = 98-0011 as the minutes of the previous meeting.

Plum summarized the WG21 meeting.

The WG20 report was deferred to the LWG meeting. Simonsen reported that two TRs have been finalized: framework for internationalization and document 10176 - guidelines for design of programming languages – that relates to the [extendedid] Appendix of the C++ Standard which is now out of date. The new C draft standard is aligned with 10176. Suggested action to align the [extendedid] Appendix of 14882 with the new WG20 TR and the ISO C CD. The WG20 sorting standard is at CD level. Enhanced locales, document 14652, features an enhanced Posix syntax for character sets, Simonsen is editor.

The WG14 report was deferred because Benito is currently in a WG14 meeting.

Koenig asked for clarification on the purpose of the WG sessions. Plum said CWG & LWG have lists of issues to work on, mainly to deal with reported defects, questions and comments in a technical manner.

The committee recessed at 9:50PST on Wednesday, 7 October, 1998.

Clamage reconvened the meeting at 13:40PST on Thursday, 8 October, 1998.

Gibbons summarized the WG’s progress. He said 48 issues had been covered, most were not defects or were very minor. A couple of large items have recommended solutions:

• Exception specifications should be part of the type system, i.e., statically checked. Myers noted that this might have a large impact on the library. Gibbons said the current situation is untenable as some parts of the Standard allow programs to compile but do the wrong thing. Ball said the previous attempt to provide type-like checking didn't work. Gibbons categorized that as a "shadow type system". Myers’ concern centers around the library specification allowing implementers to choose whether or not to provide exception specifications - this proposal would potentially make several current implementations non-conforming. Further discussion is clearly needed - papers will appear in future mailings. Spicer noted that CWG voted all in favor with a few abstentions.
• Default initialization based on Koenig's paper in the pre-meeting mailing. CWG almost unanimously in favor of Koenig's recommendation. Koenig will attempt to provide revised wording. Koenig was keen to provide early guidance to users and implementers.

Template issues will be looked at by Core WG this afternoon. Gibbons said there were several ‘extern "C"’ issues where the Standard is vague. He said Core has a proposal to clarify behavior that contains a few items that might be controversial. He said calling ‘exit()’ within a signal handler was a problem because ‘exit()’ invokes static destructors which are library functions, hence it violates the rules in the Standard. Core want to make it undefined (as WG14 have done). WG14 have also said calling ‘atexit()’ within exit processing is undefined - we should adopt this. Myers wanted clarification about static objects constructed during exit processing. Gibbons said this should be undefined too.

Dawes summarized the WG's progress. The WG have covered 70 issues so far out of 93. Most of these are minor issues that the WG closed, most at the level of typos.

Dawes said there is a paper on "namespaces for library extensions", N1166 = 98-0023 by Myers. This is preliminary work but represents what the LWG would like to progress into a TR. Wording to propose a TR should be available in Dublin. Interested parties should read this paper and contribute feedback to Myers. Myers wanted more information on what’s involved in making a TR proposal. Plum to discuss later on.

Dawes then discussed the WG20 Internationalization API proposal, N1168 = 98-0025. The sense of the LWG was to use c++std-intl to build consensus and formulate a possible C++ binding. Several possible approaches were briefly discussed in the WG but there was no agreement on the actual direction.

Kehoe summarized the WG's progress. They have draft a proposal for a New Project. They have volunteers to work on each section of the proposed report. Cygnus will host a web site for PWG containing various large documents. There are now three NP proposal documents that we need to decide on the approach:

• a combined performance and low-level programming TR
• a performance TR
• a low-level programming TR

Plum tried to explain why a single TR covering all performance and limited-resource issues was being proposed. Plum said the group felt it would be more productive to work on a single TR.

Plum said the majority of PWG wanted to split out the low-level programming TR. Myers asked for clarification that each TR needed a NP. Plum said yes. Plum said there were four NBs in support of this work with the fifth being confirmed just prior to the Plenary in August. A NP needs five NBs supporting to progress it. At the Plenary, two more NBs expressed support based on the addition of the low-level programming issue. Plum is concerned that we might lose NB support if we propose to split the TR.

Plum noted that nothing on the PWG agenda would involve Core extensions, only Library extensions. Josuttis asked whether we could have a WG21 meeting at some point this week to discuss this. Plum agreed to defer further discussion until Friday.

Plum explained the proposed process as detailed in the minutes of the Nice meeting (March '98), N1154 = 98-0011.

• The public can submit issues via comp.std.c++ or the UK panel. The moderators of comp.std.c++ or members of the UK panel can reject issues or forward them to the issue list maintainers.
• CWG and LWG can also bring forward issues directly to the list maintainers.

Then CWG and LWG work on the issues lists. Potential DRs are brought to the full committee (J16). The project editor takes the result of the J16 decision and produces the official DR list. The DR list goes forward to the Convener.

Also, NBs can create DRs that go directly to the Convener.

The Convener forwards DRs periodically to SC22. DRs are then worked on by the full committee. The results are then voted on (at the following meeting) and published. Dawes said LWG wanted to expose the issues list to the public much earlier in the process. Glassborow also spoke in favor of making the issues list very public.

The committee recessed at 15:25PST.

Core, Library and Performance WGs worked from Wednesday morning through Thursday lunchtime and Thursday afternoon onwards.

Clamage reconvened the meeting at 8:44PST on Friday 9 October 1998. Clamage said the WG reports indicate that a lot of useful work was done at this meeting. We also have the basis of the DR process agreed. There will be a brief US Tag meeting immediately after this J16 meeting closes.

4.1 Formal motions

There were no formal motions, not even to approve the WP! Laughter. Saks wanted to know if we were voting on the NP proposals. Plum said Nelson needs a sense of J16 in order to cast a vote in the WG21 meeting later. More discussion about process. Plum felt we had formal motions - the NP proposals. Corfield objected that we had not yet agreed which combination of NP proposals should go forward. Plum affirmed that he intended to act on whatever consensus was reached in WG21.

Clamage declared that we were in committee of the whole so we could conduct straw votes.

Plum felt that J16 had agreed in principle to support the result of either proposal (one combined NP or two separate NPs). It was clarified that neither of these two options was actually represented by a paper in the pre-meeting mailing.

Corfield said the only real concern is a procedural one: we know we have enough NB support for one NP but we don't know for sure whether we will get enough NB support for the portable I/O NP if we split it off on its own. Simonsen agreed and suggested our fastest and most secure way forward was to submit one NP.

Corfield spoke out strongly against trying to split the work because we have no idea whether we'd get the support for the portable I/O work whereas almost everyone actually wants to do the work. Simonsen agreed. Koenig voiced support for Simonsen’s and Corfield’s position: we go forward with one NP and think about splitting it later. Sutter agreed and said we can deal with organizational issues of the document later.

J16 straw vote indicating preference for a single combined NP or two separate NPs:

J16 straw vote indicating support for either approach:

Straw votes indicating support for each separate NP:

Nelson asked for a show of hands indication of the level of participation:

• J16 members likely to participate in performance NP work: 9.
• J16 members likely to participate in portable I/O work: 6.

Convener proposes to put forward a single combined NP:

Convener proposes to put forward wo separate NPs:

Clamage declared the committee to be back in General Session.

One action item omitted from yesterday: a request from WG14 to identify those CWG/LWG issues that might impact WG14. Dawes felt that part of the list maintenance should probably include this anyway. Simonsen volunteered to help coordinate this.

Miller asked what the status of the DR list was. Koenig said we have no confirmed DRs yet to be placed in a report. Plauger clarified that the Project Editor has the authority to declare a Defect Report but the Convener officially maintains the list.

[Note: this topic was actually covered in General Session on Wednesday, due to O’Riordan’s commitments on Friday, but is listed here for consistency with the agenda.]

O'Riordan discussed the next meeting in Dublin. More details will be in the post-meeting mailing. The meeting will be 12-16 April '99.

Miller asked that people email him copies of any documents they created here. Deadlines are 23 October 1998 for post-Santa Cruz and 23 February 1999 for pre-Ireland mailings.

Plum Hall is hosting the October 1999 meeting in Kona, Hawaii. Clamage asked people to think about hosting future meetings. Simonsen is looking into hosting a meeting in Denmark in 2000.

Plauger moved that we thank our hosts. Applause. Plum noted that Perennial should be recognized for the networking supplied.

Motion by Glassborow/Charney to adjourn:

Plum reconvened WG21 at 10:41PST Friday 9 October 1998.

Plum asked that we repeat the introductions since we have different delegates present than for the 6 October, 1998 session.

Plum explained that the purpose of this meeting was to approve forwarding the NP proposal. Kristoffersen feels it doesn't matter whether we have one or two NPs but it seems reasonable to move forward with one combined NP and later split it if necessary. Plum said we seem to have the support for the combined work even if there is not as much specific support for the portable I/O section of it. Plum said we understand that delegates’ votes here do not guarantee what the actual SC22 vote will be. Plum confirmed that we can always split a TR in the future if we need to and that there was no guarantee that the resulting TR would look similar to what is currently on the table in terms of the technical details.

Plum called for a formal vote to support forwarding a single combined performance NP:

The NBs in favor were Canada, Denmark, Germany, Ireland, Japan, UK, US.

Plum requested that we try to draw new people to the C++ standards process whose background is embedded / real-time etc.

Simonsen would like WG21 to review two FCDs from WG20. We should be prepared to discuss this in Dublin. One is on sorting, the other is on locale issues.

Glassborow feels the co-location of WG14 & WG21 has been very beneficial and wanted to extend thanks to WG14 for having us around. He also said that we should consider co-location for the year 2000 (we already have a co-located meeting in Hawaii in 1999). Glassborow felt that the co-located meetings would benefit from being in the US. Plum proposed that we thank WG14 for their excellent liaison and responsiveness. Motion accepted by acclamation.

Benito said that the co-located meeting was also useful for WG14 but their schedule for Hawaii (October 1999) was not yet fixed.

Motion to adjourn.

Plum adjourned the meeting at 11:18PST, 9 October, 1998.