ISO/IEC JTC1/SC22/WG5 N1056 To: X3J3 and WG5 From: Jerry Wagener Subject: JLW informal report of the August 1994 WG5 and X3J3 meetings Date: 28 August 94 As you know by now, unless you have been incommunicado the past week as have I, the requirements for Fortran 95 were firmed up and finalized at the WG5 meeting and the technical work for all technical additions to Fortran 95 was completed the following week by X3J3; the latter was pased subject to email ballot ratification by the full X3J3 membership, since only half the membership was present in Edinburgh - my thanks to David Muxworthy for distributing the ballots, in my absence, following the meeting. This note is my informal report on these two meetings, but first the most important thing. On Saturday following the X3J3 meeting, Jean and I leisurely made our way to a mist-enshrouded Fort William, arriving at dinner time in a downpour. The plan was to headquarter there, waiting for "the first good day" for me to try the "back way" up Scotland's highest mountain, Ben Nevis, via the Carn Mor Dearg arete (knife-edge ridge). The next day, Sunday, turned out to be that day. By morning the clouds had lifted and, as it turned out, stayed off the summit for the entire day. A local climber said they don't get more than "five days like this" in a year - I even got a sunburn. The route was great, giving spectacular views of the famous Ben Nevis cliffs most of the way. Total ascent via this route is 1394 meters, just short of a "metric mile". The arete turned out to be more rock scramble than high exposure - time consuming but not particularly dangerous (but I wouldn't advise tackling those moss & lichen covered rocks under wet conditions). The next day the mist had again claimed Ben Nevis and we departed Fort William as we had arrived, in the rain. Two weeks earlier, at its opening session, WG5 received the X3j3 report on the status of the almost 20 tentative requirements for Fortran 95, established at the 1993 WG5 meeting. Most of those requirements had been implemented. Of the remaining items, WG5 accepted X3J3's recommendation on several to defer to Fortran 2000, but that left three requirements unresolved: allocatable structure components, condition handling, and parameterized derived types. In addition, the US had proposed two new requirements for Fortran 95. It was clear that in order to make the April 1995 CD (Committee Draft standard) delivery schedule, X3J3 had to complete all new technical work during its meeting the following week, and spend the intervening time until April on polishing the CD document (the 007 document, in X3J3 terms). WG5 was firm that maintaining the schedule was far more important than the specific technical content of Fortran 95. Therefore no new requirements were established, beyond those already identified, and ad-hoc task forces were commissioned to investigate the three unresolved items. The task forces on allocatable components and condition handling reported possible resolution of the technical isssues concerning those requirements and recommended retention of these items as requirements for Fortran 95. the task force on parameterized derived types did not resolve the related technical issues and recommended that this requirement be deferred until Fortran 2000 and be integrated with the work on object-oriented technology. WG5 accepted these recommendations and prepared the final list of Fortran 95 requirements accordingly. By the end of the week it was clear that WG5 should perhaps leave the door open to two other technical issues, should X3J3 be able to complete them for Fortran 95. One was automatic deallocation of allocatable arrays - an issue that arose in connection with (but broader than) allocatable components. the other was one of the US proposed requirements, user-defined elemental procedures, because of its potential importance in high performance computing. In the end the WG5 resolution on Fortran 95 requirements, though not including either of these items, was worded such that either or both could be included in Fortran 95 should X3j3 be able to incorporate them into the CD. One of the objectives of this WG5 meeting was to establish initial requirements for Fortran 2000. Because processing the work of the three technical task forces on Fortran 95 requirements was of higher priority, only general directions for Fortran 2000 were established. These took the form of three "levels". Of highest priority was to maintain Fortran's leadership for scientific and engineering applications requiring high performance. Next was a two-pronged investigation of object-oriented technology and Fortran interoperability with other environments; task forces were established for each of these, to report findings back to WG5 in early 1995. The third category of Fortran 2000 requirements were miscellaneous improvements to the language, though no specific items were identified for this category at this meeting. At the X3J3 meeting the following week, the principal objectives were to (1) complete the technical work on Fortran 95 requirements, (2) continue with the Fortran 90 interpretation processing, and (3) clarify the editorial processes for preparing the CD document. There were a total of 15 attendees at this meeting, with a total of 12 votes. (There are 25 members of the committee, and a quorem is 9.) Regarding objective (1), the four items for which there were reasonably mature proposals to start with were allocatable components (94-269r2), condition handling (94-258r4), user-defined elemental procedures (94-245r3), and automatic deallocation of allocatable arrays (94-270r3). The technical subgroups focused on these items and in the end all four were approved for the 009. None were approved unanimously, however, and because of the 50%-ish attendance they were approved subject to a full committee letter (email) ballot. (The ballots have already been distributed.) The main purpose of these ballots is to identify any technical "showstoppers" that were not uncovered during the meeting. because of the tight schedule prior to the next X3J3 meeting, two weeks is allocated for the email ballot for those items for which all technical concerns seemed to have been resolved (allocatable components, user-defined elemental procedures, and automatic deallocation). At the last minute a potential problem was identified, and potentially resolved, with the condition handling item; because of this condition handling will be balloted for 30 days. Regarding objective (2), at this meeting a total of 185 requests had been submitted for Fortran 90 interpretation, of which 46 had not yet been resolved by X3J3. During the meeting draft responses were approved for 21 of the 46 unresolved items; these draft responses will constitute the post-meeting ballot on interpretations. That leaves 25 unresolved items, work on which will continue at the next meeting. While interpretation processing will continue, the main objective for the next meeting, in Boston in November, will be refining the 007 document into CD form. The content of the 009 document was completed at this meeting, the 009 will be archived, and its contents will be integrated into the 007 between meetings. The latest 007 will be the base document for proposals at the next meeting, the bulk of which is expected to be fixing any problems with the integration of the 009 material; in addition, work on the editorial items of the 008 will continue. The projected 007 schedule (94-286) calls for revisions in mid-September, mid-October, mid-December, mid-February, and late May, the last of which will be the CD; at least two further revisions are anticipated before it becomes the Fortran 95 standard.