Minutes of January 2005 Meeting

Austin/239
Minutes of the 9th Plenary Meeting of the Austin Group
11-13 January, 2005
Reading, UK, Hosted by The Open Group.

Attendees

Name	Affiliation	Role
Andrew Josey	The Open Group	Chair
Nick Stoughton	Usenix	SC 22 OR, Secretary
Don Cragun	Sun Microsystems	IEEE OR
Keld Simonsen	DKUUG (teleconference, part)
Larry Dwyer	HP (teleconference, full)
Mark Brown	IBM (teleconference, full)	The Open Group OR
Mats Wichmann	Intel
Cathy Fox	The Open Group (part)
Geoff Clare	The Open Group
Joanna Farley	Sun Microsystems
Stephe Walli	USENIX
Joe Gwinn	Raytheon,SSWG-RT WG
Stuart Anderson	Free Standards Group (teleconference, part)

Andrew Josey called the ninth meeting (a.k.a. Austin/M10, since this counting includes a teleconference) of the Austin Group to order at 9:03 am Tuesday January 11th at The Open Group's office in Reading, Berkshire.

Meeting Goals

The goal of this meeting is to

• Generate the work plan for 2005-2007
• Discuss proposals for new work items
• Resolution of the open aardvark input to date.

1. Coffee and Cakes
2. Administrivia
   • Introductions

     Stephe Walli and Mats Wichmann were welcomed.

   • Approval of Agenda

     The agenda was approved as published, with the addition of a discussion on Memory Synchronization (New Business).

   • Approval of Minutes

     Minutes of the last teleconference ( Austin/232, December 16 2004) were reviewed. Approved with no objections.

   • Review of Action Items

     ACTION AI-2005-01-01: Don Cragun to present paper on pathname resolution. OPEN.
     ACTION AI-2005-01-02: Larry Dwyer to present proposal on system() and threads. OPEN.
     ACTION AI-2005-01-03: Joerg Schilling to address issues in XCU ERN 1 (pax) OPEN.

   • Review of Consent List

     SD2 updated. Added 37, reversing Item 7 (option _POSIX_PII) since although we agreed at one point in time that this option should be present, it has not been implemented (networking subsequently became part of the mandatory base functionality), and is now overcome by events.

   • Review of Issues List

     SD1 - no updates.

3. Procedures Update
   • JPC Procedures
     No updates to report.
   • Committee Draft development procedures. See item 5 below.

   No updates.

4. Identification of Organizational Reps

   see attendance list above. WG 15 has been disbanded, but the representation from ISO/IEC continues, direct from SC 22, with Nick Stoughton as the Organizational Representative (Keld Simonsen is the alternate). Currently, the representative is the "interim" rep, until a recommendation is received from the newly formed POSIX Advisory Group for a permanent rep. The PAG met yesterday, and recommended that the two interim reps be made permanent. This recommendation has yet to be acted on by SC 22.

5. Status Reports
   • Austin Group Status - presentation by Andrew Josey (Austin/234).
     Balloting Status: The Open Group, IEEE, ISO

     TC 1 and TC 2 have both been approved and published by all three organizations.

     Spiral bound copy of 2004 edition available: 7 volumes ... approximately 9 inches of shelf space required! Document on-line in HTML, and Guide Book with PDF on CD-ROM.

     ISO have chosen not to publish the 2004 edition as a separate edition. Therefore, printed copies of the 2004 edition only carry the two logos. However, ISO have approved TC2, just chosen to publish it as a standalone edit set (i.e. not an integrated document). ISO call the second technical corrigenda TC1!

     Approval dates for TC2:

     1. The Open Group December 2003
     2. IEEE February 2004
     3. ISO September 2004

   • Committee Draft development procedures.
     Geoff Clare asked how to handle an enhancement request to add a new option to a utility. A well formed aardvark should suffice.

   • Organizational Reps Status - No issues.
   • POSIX Status
     The System Services Realtime group (chaired by Joe Gwinn) met July 2004. POSIX.13 and POSIX.26 (device control) are approved (at the December 2003 IEEE Standards Board meeting). IEEE published these in 2004. New work items are being discussed, but things have been fairly quiet for the last six months. The device control specification could be considered by the AG for future inclusion in a revision of the core should the PASC OR submit it and the other recommended items from Austin/112r1 be met.

     The system services realtime working group is the only active group remaining within PASC, and the POSIX.13 profile amendments are the only documents still under development outside the AG. Most recent status report is in the presentation at Austin/235.
     ACTION AI-2005-01-04: AJ to prepare a slide showing the timeline of the POSIX standards based on one within Austin/235

   • IEEE/The Open Group Status
     The Open Group approved the standard September 12th 2001. The IEEE approved the standard December 6th 2001. Subsequently they have worked together to establish a certification program for POSIX (see Austin/237).
   • C99 Status, C99 TC2
     Nick reported that WG 14 is working on TRs for "Security" (see below), Embedded C, new character data types. TC 2 was published in November 2004, and we should align with it. ACTION AI-2005-01-05: AJ to prepare draft document on impact of C99 TC2 on POSIX.
   • Networking Status (XTI/Sockets)
     IETF draft RFC draft-itojun-ipv6-getnameinfo-multiproto-02.txt was submitted to the futures mail list for our consideration. Andrew Gollan has left Sun, and is no longer the liaison for networking issues. ACTION AI-2005-01-06: AJ to contact Jack McCann at HP for views on the getnameinfo extensions document, and for any other networking issues that may concern us.
   • Joint Procedures Committee Status
     Nothing new.
6. Old business
   Utility Syntax Guidelines

   Long running discussion on e-mail list on this subject. A number of utilities that had "obsolescent" alternative syntaxes had those alternates removed during the revision. This is summarized in austin-group-l message 7269.
   ACTION AI-2005-01-07: AJ to update AI-027 notes to the Editor and recirculate.
   Going through this list utility by utility:

   • set if the first argument is '-', the results are unspecified.
   • awk skip...this is different.
   • ed append to the first para in "Options" the following: "except the unspecified usage of '-'."
     Add to DESCRIPTION "If an operand is '-' the results are unspecified."
     Delete the last sentence in the section on file under OPERANDS.
   • env append to the first para in "Options" the following: "except for the unspecified usage of '-'."
     Add to DESCRIPTION "If the first argument is '-' the results are unspecified."
   • find no change needed.
   • grep no change needed.
   • head add to rationale "Earlier versions of this standard allowed a -number option. This form is no longer specified by this standard but may be present in some implementations."
   • join add to the Rationale "Earlier versions of this standard allowed -j, -j1, and -j2 options, and a form of the -o option that allowed the list option argument to be multiple arguments. These forms are no longer specified by this standard but may be present in some implementations."
   • kill no change required.
   • sort some implementations allow the operands and options to be intermixed, in violation of the utility syntax guidelines. Add to OPTIONS section on conformance to USG "except Guideline 9". Add to the RATIONALE, "Earlier versions of this standard allow the -o option to appear after operands. Historical practice allowed all options to interspersed with operands. This version of the standard allows implementations to accept options after operands, but conforming applications should not use this form.

     Earlier versions of the standard also allowed -n and +m. These options are no longer specified by this standard, but may be present in some implementations." Also add to OPTIONS "In addition, '+' may be recognized as an option delimiter as well as '-'."

   • tail Add to the Rationale: "Earlier versions of this standard supported the following forms in the Synopsis section:

     tail -[number][b|c|l][f] [file]
     tail +[number][b|c|l][f] [file]
     

     These forms are no longer specified by this standard, but may be present in some implementations." Also add to the OPTIONS "except that '+' may be recognized as an option delimiter as well as '-'."
   • touch see XCU ERN 45 (already accepted aardvark).
   • tty no change required.
   • uniq in the OPTIONS section add "except that '+' may be recognized as an option delimiter as well as '-'." Add to Rationale "Earlier versions of the standard allowed -n and +m. These options are no longer specified by this standard, but may be present in some implementations."
   • ex add to Description "If an operand is '-', the results are unspecified." Add to OPTION "except for the unspecified usage of '-', and that '+' may be recognized as an option delimiter as well as '-'." Add to Rationale "The +command option is no longer specified by this standard but may be present in some implementations."
   • expand add to the Rationale "Earlier versions of this standard allowed the following form in the Synopsis section:

     expand [-tabstop] [-tab1,tab2,...,tabn] [file ...]
     

     This form is no longer specified by this standard but may be present in some implementations."
   • more Add in the OPTIONS section to the end of the 1st para at the end of 24927 "except that '+' may be recognized as an option delimiter as well as '-'." In the RATIONALE section on page 653 add to the end of line 25536 (line starting "The -p (position)...") "The +command option is no longer specified by this standard but may be present in some implementations."
   • newgrp Add a new paragraph after 25589 "If the first argument is '-', the results are unspecified." Append to end of first paragraph of OPTIONS (line 25600) "except for the unspecified usage of '-'."
   • nice Add to the end of RATIONALE for nice as a new para : "Earlier versions of this standard allowed a -increment option. This form is no longer specified by this standard but may be present in some implementations."
   • renice Change the first paragraph in OPTIONS from "The renice utility shall conform to the Base Definitions volume of IEEE Std 1003.1-2001, Section 12.2, Utility Syntax Guidelines." to "The renice utility shall conform to the Base Definitions volume of IEEE Std 1003.1-2001, Section 12.2, Utility Syntax Guidelines, except for guideline 9." In the OPTIONS section for -g, -p -u change from "Interpret all operands ..." to "Interpret the following operands ..." Add to RATIONALE "Earlier versions of this standard allowed the following forms in the SYNOPSIS section:

     	renice nice_value[-p] pid...[-g gid...][-p pid...][-u user...]
     	renice nice_value -g gid...[-g gid...][-p pid...][-u user...]
     	renice nice_value -u user...[-g gid...][-p pid...][-u user...]
     

     These forms are no longer specified by this standard but may be present in some implementations."
   • split Add new paragraph to the end of the RATIONALE for split: "Earlier versions of this standard allowed a -line_count option. This form is no longer specified by this standard but may be present in some implementations."
   • strings Add a new paragraph at the end of DESCRIPTION "If the first argument is '-', the results are unspecified." Append to end of first paragraph of OPTIONS (line 34166) "except for the unspecified usage of '-'." Add to end of existing 1st paragraph in the RATIONALE for strings: (line 34235) "These options are no longer specified by this standard but may be present in some implementations."
   • vi Append to end of first paragraph of OPTIONS (line 38101) "except that '+' may be recognized as an option delimiter as well as '-'."
   • lex no change required.

   Work on USG related aardvark: XBD ERN 21, 22. XRAT ERN 2, 3, 4, 5, 6. All closed.
   ACTION AI-2005-01-08: AJ update the aardvark reports.

   Memory Synchronization

   There were not sufficient experts in the meeting to progress this issue. Nick Stoughton reported that ISO JTC 1/SC 22/WG 21 (C++) are considering developing new interfaces in the C++ language "because the POSIX model is broken". We should encourage all experts in this area to be prepared to discuss these issues (both the aardvark and possibly the WG21 proposal) at the next conference call.

7. New business

   • POSIX LSB Differences

     ACTION AI-2005-01-09: Andrew Josey & Mats Wichmann to propose a time and place for a joint Austin group/LSB Workgroup meeting, to be held as soon as conveniently possible.

     The group worked through the draft TR on differences (see Austin/236) and categorized the issues as follows:

     1. Errno differences
        These include ENXIO v ENODEV, ENOTSUP v EOPNOTSUP, EIDSIR v EPERM. Although these each have different potential resolutions, they should all be dealt with at a joint meeting.
     2. Substantial Interface differences
        Includes fcntl, kill and strerror_r. To be handled by a joint meeting. Both specs could change to accommodate diffs. LSB could have a new invented function to provide POSIX behavior, and use the right one if POSIX_C_SOURCE or _XOPEN_SOURCE is defined.
     3. POSIX Interpretation
        Interp request filed against POSIX, POSIX likely to change to accommodate LSB (strptime).
     4. LSB Extensions
        These are not conflicts, although listed as such. Includes waitpid, ioctl, bc, ipcrm and od.
     5. LSB Issues
        Some things just need more investigation (they may not really be conflicts): opterr, optind, and optopt. Others need investigation to figure out what to do, but both standards might be able to accommodate changes to support the other (at -r/-d). Other changes are in conflict, and should be investigated by LSB for further resolution. These include at, batch, crontab, df, chgrp, chown, cut, echo, renice, more, newgrp, xargs, file, fuser, ipcs, ls.
     6. Internationalization
        awk, find, grep, sed. Need LSB to address these. Geoff Clare reported that there are additional conflicts for the find utility that are not in the TR. He will email details to Nick Stoughton.

   • Aardvark

     XBD ERN 19 - Environment Variables discussion. The principle issue is that several implementations use an environment variable (or variables) to indicate a conforming environment (e.g. POSIXLY_CORRECT). This is actually prohibited by the standard as written. Proposal from Mark Brown for changes to allow this. The first action of Mark's proposal was dropped, but the rest accepted after lengthy debate. This is an interp. Andrew preparing draft of interp for review.

     XBD ERN 20 - withdrawn by submitter (Geoff Clare)

     XBD ERN 28 - Accept

     XBD ERN 36 - looks editorial. The proposed change would match the text in 1003.2-1992. Also RE_DUP_MAX applies to EREs as well as BREs. Editor has marked the aardvark.

     XBD ERN 3 - previously marked ACCEPT -- should be ACCEPT AS MARKED.

     XBD ERN 4 - Previously marked ACCEPT AS MARKED, but the suggested change was unclear. The action was clarified. (change "do the [EMFILE] change as suggested, note that the exact changes still need enumeration." to "Go with option 1, do the [EMFILE] change as suggested, note that the exact changes still need enumeration."

     XCU ERN 14 - seems like a worthwhile idea, but no text submitted. Asked submitter for more details.

     XCU ERN 22 & 23 - leave, pending input from Don Cragun.

     XCU ERN 48 - Accept As Marked (add the text provided).

     XCU ERN 49 - Accept as marked (see aardvark for details). This is an interp. The standard is ambiguous in the first part, and the second case is a defect.

     XSH ERN 12 - previously marked REJECT. Change to AAM ... see aardvark for further info on new words. Could be a TC.

     XSH ERN 29 - hold till the next teleconference.

     XSH ERN 52 - discussions between Loic and Dave Butenhof still ongoing here.

     XSH ERN 63 - Accept as marked ... see aardvark for detailed edits. This is an interp.

   • Open Interpretations

     AI 002 - fflush and fd interactions. ACTION AI-2005-01-10: AJ to find 1988 wording related to fflush() and fclose() here and update the proposed words, and recirc. AI002.

     AI 016 - Pathname resolution, awaiting Don Cragun input.

     AI 029 - popen("z"). Change "undefined" to "unspecified". Note that passing a null pointer remains undefined behavior because the standard says that the mode argument is a string.

   • Responses to Draft TR from WG14: "Security" Library Extensions

     WG 14 have submitted a draft of this TR for CD Registration, and have asked AG for review and comment.

     The document is titled "Specification for Secure C Library Functions" but does not cover anything related to security. It provides bound checked string functions and other interfaces intended to reduce programmer error. The AG respectfully suggest that the title should be something more like "Enhanced Library Interfaces", or similar. Security can be divided into six basic requirements, or tenets, that help ensure data confidentiality, integrity, and availability. The six security tenets are:

     • Identification. This deals with user names and how users identify themselves to the system.
     • Authentication. This deals with passwords, smart cards, biometrics, and so on. Authentication is how users prove to the system that they are who they claim to be.
     • Access control (also called authorization). This deals with access and the privileges granted to users so that they may perform certain functions on the system.
     • Confidentiality. This deals with encryption. Confidentiality mechanisms ensure that only authorized people can see data stored on or traveling across the network.
     • Integrity. This deals with checksums and digital signatures. Integrity mechanisms ensure that data is not garbled, lost, or changed when traveling across the network.
     • Non-repudiation. This is a means of providing proof of data transmission or receipt so that the occurrence of a transaction cannot later be denied.
     This TR deals only with integrity, and only a little of that.

     Namespace pollution issues ... should not be implementation defined if __STDC_WANT_SECURE_LIB is undefined. P4 para3. The undefined case should be the same as defined as 0.

     We should continue to review this document offline. Comments to austin-group-l. Nick Stoughton will deliver the comments to the next meeting of WG14. Mailing deadline (for WG14) is 7 March 2005. Our deadline should be one or two days before that. ACTION AI-2005-01-11: All to review WG14 Security TR and report to austin-group-l by 2005-3-1.

   • New APIs Strawman Doc

     Need to continue to monitor developments here. Base Working Group currently own the action here. Some overlap with WG14 security TR.

   • Access Control Lists

     There was a problem that ACLs vacillated between even and odd numbered drafts of .1e/.2c, and some implementations do it one way, others the other. There are many implementations with something in this space. Should ACLs be restrictive or permissive? This may be a split in directions. ACTION AI-2005-01-12: Larry Dwyer and Nick Stoughton to further investigate ACLs and propose next steps.

   • POSIX Certification Status Report

     Joint IEEE Open Group effort to certify POSIX implementations. Andrew presented a status report on certification (Austin/237), supplemented by material in Austin/238. Test suite covering all 1003.1 2003 mandatory interfaces (including shell). Developing new tests for 1003.13 profiles - coming soon!

   • SD/5 Review

     Remove XBD ERN 8 from here -- we subsequently rejected this change.

     Remove XSH ERN 29 from here -- still open.

   • Work Plan Update

     AGREED Need to start a revision by December 2006. (OR vote, add to SD/2)

     Would want to have PAR approved by September 2006. Other groups (e.g. IEEE, ISO, OGTGbase) have until then to finalize specs for consideration for inclusion in the revision. We should have a notification of all incoming revision items by January 2006. Procedures already in place require input to be approved specs. See Austin/112r1.

8. Consent List - Updated

9. Issues List - no updates

10. Action Item Review
    1. ACTION AI-2005-01-01: Don Cragun to present paper on pathname resolution. OPEN.
    2. ACTION AI-2005-01-02: Larry Dwyer to present proposal on system() and threads. OPEN.
    3. ACTION AI-2005-01-03: Joerg Schilling to address issues in XCU ERN 1 (pax) OPEN.
    4. ACTION AI-2005-01-04: AJ to prepare a slide showing the timeline of the POSIX standards based on one within Austin/235
    5. ACTION AI-2005-01-05: AJ to prepare draft document on impact of C99 TC2 on POSIX.
    6. ACTION AI-2005-01-06: AJ to contact Jack McCann at HP for views on getnameinfo document, and for any other networking issues that may concern us.
    7. ACTION AI-2005-01-07: AJ to update AI-027 notes to the Editor and recirculate.
    8. ACTION AI-2005-01-08: AJ update the aardvark reports.
    9. ACTION AI-2005-01-09 Andrew Josey & Mats Wichmann to propose a time and place for a joint Austin group/LSB Workgroup meeting to consider LSB/POSIX conflicts, to be held as soon as conveniently possible.
    10. ACTION AI-2005-01-10 AJ to find 1988 wording related to fflush() and fclose() here and update the proposed words, and recirculate AI002.
    11. ACTION AI-2005-01-11 All to review WG14 Security TR (N1093.pdf) and report to austin-group-l by 2005-03-01.
    12. ACTION AI-2005-01-12 Larry Dwyer and Nick Stoughton to further investigate ACLs and propose next steps.
11. Draft & Meeting Schedule

    The next teleconference is January 27 at the regular timeslot. We do not expect the next face to face meeting until January 2006, but will review this periodically.

12. AOB
    1. All members of the Austin Group were saddened to hear of the death of Finnbarr Murphy's son, and send heartfelt condolences to him and his family at this time.
    2. All members of the Austin Group wish to thank The Open Group for hosting this meeting and taking care of our needs so dutifully.

    The meeting adjourned at 3.16pm Thursday.

    ---

    If you have comments or additions relating to the correctness of the Minutes please send them to The Austin Group Chair. If your comments relate to decisions regarding the disposition of the aardvark please refer first to the Aardvark reports, and then if you still have a comment please send it to the main reflectors so that all can see the issue.