BSI IST/5/-/15 (Posix Panel) Unconfirmed Minutes of Meeting 30-Aug-91 Attendees: Dave Cannon University of Exeter * Derek Jones Knowledge Software Richard Pearson BSI Eric Davie CCTA * David Brownbridge Praxis Systems Plc Brian Meek King's College, London * John Merrells BSI QA * Dominic Dunlop University of Oxford * Martin Kirk X/Open Apologies: Vince Rich Encore Hugh Davis John Dawes * - These people received no mailing: see below. 0. Agenda: The agenda was agreed as circulated. The meeting began with a presentation by Brian Meek on the work of IST/5/-/11, known as the Bindings Panel, with particular reference to its current work on CLID, the Common Language Independent Datatypes standard. CLID Presentation: IST/5/-/11 was established in about 1982 and has fostered a number of standard projects, mostly concerned with cross- language issues. The most recently acquired are - LCAS - Language Compatible Arithmetic Standard CLID - Common Language Independent Datatypes, and CLIPcm - Common Language Independent Procedure calling mechanism. (here the 'mechanism' is widely misinterpreted as an implementation mechanism, whereas the scope of the project specifically says that the document describes a language mechanism for invoking procedures). Obviously CLIP, describing as it must the form of parameters used in invoking Procedures, relies heavily on the definition of datatypes for those parameters, hence CLID. The major initial effort therefore went into the CLID standard. Among the required datatypes are the arithmetic datatypes, including the 'real' (or 'float') datatype. It was felt that there was a need to define arithmetic operations as a standard to underpin the CLID work, and this has led to the LCAS draft work (for Language Compatible Arithmetic Standard) possibly to be renamed CLIA (for Common Language Independent Arithmetic). WG11 is in a sense managing these in sub-groups, though there is an obvious inter- relationship which prevents total partitioning. A background document is due for revision at the next WG11 meeting and will be circulated to WG15, describing the work in progress. CLIP was originally intended to be fast-tracked through ISO, but collided with the RPC work in SC21. The two efforts are now proceding in close liaison. All the drafts use a common interface definition notation, which is 'enhanced ASN-1'. ASN-1 itself is not adequate for this purpose. Both CLID and CLIP are close to going to CD; CLID is already technically at CD ballot, but this was premature. CLID has a set of primitive datatypes, including NULL and 'undefined', which are stable and have been so for some time. The CLID standard goes on to define a full set of datatypes, including the more complex ones (the maximalist approach) rather than defining, or leaving to be defined, other types in terms of these (the minimalist approach). The CLID standard does not yet address the definition of datatypes of the form of the newer object-oriented styles. This may come in as part of a revision although there is some pressure to introduce them now. It is the intention of the standard to provide a two-way mapping between the datatypes of the CLID and the datatypes of the CLIP for each current language which needs to use it. CLID is aimed at making multi-language products more portable. It is a basic requirement of the RPC work, and can be regarded as an 'enabling' standard. 1. Minutes The minutes of the previous meeting were accepted with an amendment to Section 4. (Liaisons): this should read 'MK will attend IST/-/2 as official liaison'. A number of those present at the meeting reported that they had received no mailing prior to the meeting (marked '*' on the above list). The secretary is asked to check the mailing list. Action: Vince Rich. 2. Matters Arising Actions from the May meeting were summarised: Email distribution service for WG15 and Panel: MK's description is attached. Action closed. Corrupt copies of documents P122 and P123: Corrected copies have been distributed. Action closed. Potential Cobol binding to Posix: MK reported that there was no immediate interest at X/Open. DD also asked X/Open on behalf of EurOpen, (formerly the European Unix User Group) - the response was that if EurOpen could fund the exercise X/Open would help. There does not appear to be a great pressure from users to do this. Open Action: Don Hay (ICL), Paul Cheshire (IBM). Andrew Walker (X/Open) attended the SGFS meeting, Berlin, as the official WG15 liaison. Action closed. UK volunteer for the RG on Coordination of Profile Activities: Eric Davie suggested Don Folland may be available. [Post meeting note: DF has volunteered]. Action closed. Chair to advise Kevin Murphy of his acceptance as liaison from WG15 to SC21/WG1. Done. Action closed. Request to UK to host WG15 meeting in Spring '93 or either meeting 1994: Authorisation from IST/5 was given. Open Action: Panel members to seek potential hosts to provide meeting rooms or cover costs for same. MK to check on a possible offer from X/Open. (Contact Chair for details) Chair to post Posix networking group summaries to wg15-uk email list. Done. Action closed. 3. Progress of Posix Standards a) ISO/IEC JTC1/SC22/WG15 progress: ISO/IEC JTC1 had circulated a set of five New work item Proposals (JTC1 N1374 to N1378) for consideration by the member bodies. The Panel agreed to recommend 'yes' votes, which were sent from Hampden House to the IST-5 secretary for return to ISO by 20-September-1991. It was noted that the IEEE had released 'Posix: Shell & Utilities: User Portability Extension 1003.2a' Draft 6 to ISO WG15 for review and comment. The draft is ~320 pages and will ONLY be distributed to Panel members who request a copy. Copies can be obtained from either the Chairman or the Secretary. Action: Chair to advertise the Draft via email. [Post-meeting note: Done.] Action: Interested members to request copies of the UPI Draft. The next meeting of ISO/IEC JTC1/SC22/WG15 will be in Stockholm, Sweden starting with the Rapporteur groups on Monday 4th November and continuing through to Friday 8th November. The following Panel members will be attending and are recommended to IST/5 as the UK delegation: Martin Kirk (as Secretary) Dominic Dunlop Don Folland Dan Chacon (RG only, probably) Andrew Walker (possibly) Dave Cannon (PUKE) b) IEEE TCOS progress: The last meeting of the IEEE Posix groups was in Santa Clara, July 8 to 12. The following is a summary of the Groups by those who attended: Ad-hoc Group on Profiles: Beginning to develop some rules for developing profiles. Ad-hoc Group on Architecture & Framework: Will become an official [as opposed to ad-hoc] group. 1003.0: Ready for mock ballot after the October meeting. Deadline for responses to the ballot: 6 January. 1003.3 [Conformance] The 1003.3 base document is now an approved IEEE standard, published April 1991. .3.1 Recirculated in ballot, May '91. 80% approval in returns so far. .3.2 Current draft is 12. An 'Accredited Posix Testing Labs' list has been issued by NIST. NCC is on the list. A couple of implementations [A/UX is one] have already been accredited. 1003.4: [Real Time] Still in balloting. First recirculation is now complete. D11 will be issued to ballot group in September. The major changes are in the IPC section. IPC has been passed to the 1003.12 group, but .4 retains an IPC 'stub' chapter which is much simplified. It is expected that D12 will become the standard. Action: Chair to send a copy of 1003.4 D12 when available to Dominic Dunlop. Sychronisation primitives currently differ in the .4 and .4a draft standards, but there is an intention to harmonise them. It is hoped that the 1003.12 IPC section will be a superset of the 1003.4 IPC section. .4a [Threads] This is in the ballot resolution process. The 'Priority Ceiling Protocol' caused most objections. This will be replaced by 'Priority Ceiling Protocol Emulation' - a vast simplification of PCP, but allegedly with very little degradation in performance. The WG approved the replacement. The 1003.14 group are not happy, however. [MultiProcessing AEP] .4b: Some progress; timeouts will now apply to all .4 & .4a facilities - ie calls won't necessarily block forever. ISO & interrupts: the ISO comments were passed to a small WG which couldn't decide what to do. There is a new proposal on process generation to provide an alternative to fork() and exec(). 1003.4b is not yet in draft. [Some considerable unease has been expressed about how the .18 profile may affect the direction & content of other profiles, particularly the Real Time AEPs...] 1003.5 [Ada Binding] The PAR has been approved for Real Time Binding. There will be no .5a group. A new Chair is approved from tomorrow [Jim Longeurs?]. Some preliminary work has been done on the RTB. Scope of RTB will be within .4, .4a & .4b. Ada bindings to Real Time will NOT include any functionality which is already present in the language. This will not deny the possibility of an Ada program calling a 'c' program to do those things. 1003.7: [System Administration] Working on Print mgt, User mgt and Software mgt. The group has identified areas of work which can be dealt with as separate entities, & is dealing with a small number of them in parallel at any one time. 1003.8: [TFA] Has reviewed its test assertions and will go to ballot on its D5. They are postponing the Data Stream Encoding part of the PAR/Standard. 1003.12: [PII] First draft in next mailing. 1003.13 [Real Time AEP] Now have 4 AEPs. The group has subdivided to deal with these. Each AEP lists facilities it needs of the Posix Standard. [Here meaning .1, .2 and .4 & .4a]. 1003.13 will be later revised to include other Posix standards as they emerge. Each AEP may require certain performance metrics and may also need certain [currently non- existent] device control. No progress has been made on the LIS or on Test Assertions. 1003.17: [Directory Services] Document [D?] to go out to mock ballot. [Sept?] 1201.1 [Windows Toolkit API] 1201.2 [Driveability] During the week a formal request was made to approve PARs for both GUIs [Motif + Openlook] or reasons why not. The Ad Hoc committee was to respond to CS SEC with recommendations for the formal reply. Late in the evening on the 11th both proposals were rejected [for reasons of maturity of the base work, etc]. Some objections were terminal - ie they cannot be resolved: eg: "No standard is better than two standards" The rejection of the GUI PARs resulted in a motion to withdraw support for the 1201 groups. [Robert Bismuth - Jim Isaak's chief at DEC]. Paul Gambrel proposed this be passed to SEC. It was decided to leave the suggestions on the table till the end of the meeting, when it was proposed that PMC be authorised to look at the problem. [1201 is up for review by the Project Management Committee anyway] 1238.0: [OSI Interfaces] D2 reviewed. Mock ballot after next meeting, with luck. First ballot in 2Q92 with only some of the Test Assertions. Final ballot 1Q93. The next meeting of the IEEE Posix groups will be in Parsippany, New Jersey, 21-25 October 1991. (Close to the IEEE HQ). This meeting will attempt to rationalise the disparate and sometimes mutually-exclusive standards coming out of groups such as Security and TFA. 4. Rapporteur Group Reports a) Security: (KM) A report was circulated in the mailing. (P126) b) Verification Report: None. c) Internationalisation: (DD) The UniCode/ISO Wars: Consensus is being sought. One solution may be for UniCode to fill one of the frames of the proposed ISO standard. Locales: The current Posix model maps on to a model of computing which doesn't fit the philosophy of Windowing systems, where the window and its contents (may) need different locales. X/Open and the X-Consortium have suggested the concept of 'distributed locales' to solve this; at the moment the feeling of the group is that they want to pin down 'standard' locales before moving on to anything else. d) c++: A c++ meeting is scheduled to take place after the end of the next c Panel meeting. The UK are hosting an International meeting of c++ next March. 5. Liaisons a) There will be an SGFS sub group meeting re Profiles for OSE, from October 30 to November 1 at the EWOS Central Secretariat, Brussels. Prospective attendees should advise by September 15th. b) MK attended the last IST/-/2 meeting (the Functional Standards Coordination meeting) as a representative of the Posix Panel. Another meeting will be held on the 23 September, when they will be discussing 'The Way Ahead' document, (SGFS:N402, IST/-/2:N398), on which they would like feedback. This looks at the issues of extending Profiles into Open Systems environments rather than just into OSI. This appears to parallel work going on in both the IEEE and EWOS, in the CAE group. MK needs guidance on an action item stating: "there is a need to identify sources of input for CAE areas. A UK response is required regarding SC22 activities in CAE. EWOS EG CAE has distributed a report to be approved at the EWOS TA in September. A composite view was required which would indicate how this work will progress. IST/5/-/15 was instructed to make comments on this issue to IST/5." There is no record of this instruction having reached the Panel through IST channels: as a response the Panel asked MK to pass a list of known Posix documents and Draft standards relating to Profile work back to IST/-/2, including: IST/-/2:N382 P1003.18 D3: Posix PEP 6. Any Other Business None. The meeting ended at: 12:15 Next meeting: 29-November-1991 10:30 Hampden House