From Miles.Ellis@etrc.ox.ac.uk Tue Jul 30 20:00:52 1996 Received: from oxmail3.ox.ac.uk ([163.1.2.9]) by dkuug.dk (8.6.12/8.6.12) with SMTP id UAA24283 for ; Tue, 30 Jul 1996 20:00:43 +0200 Received: from vax.ox.ac.uk (actually host vax) by oxmail3 with SMTP (PP); Tue, 30 Jul 1996 18:59:55 +0100 Received: from 163.1.85.1 by vax.ox.ac.uk (MX V4.2 VAX) with SMTP; Tue, 30 Jul 1996 18:59:51 +0100 X-Sender: MELLIS@vax.ox.ac.uk Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Tue, 30 Jul 1996 19:02:37 +0100 To: sc22wg5 From: Miles.Ellis@etrc.ox.ac.uk (Miles Ellis) Subject: Convenor's informal report on WG5's Dresden meeting Note: This short report is my personal view of the major issues discussed and decided during WG5's meeting in Dresden. The formal decisions made are all recorded in the Resolutions, while a more detailed account of the meeting will be available in due course in the form of the Minutes of the meeting. =========================================================================== WG5 had a successful meeting in Dresden last week when we resolved a number of important issues, and set the direction for Fortran 2000. The meeting was also notable for having seven countries represented (Austria, Canada, Germany, Japan, Sweden, UK and USA) - more than for some years. One result of this increased attendance was that we were able to (provisionally) identify hosts for the next four meetings. One of the most positive aspects of the recent work on TRs and other technical issues by email has been a growing sense of collaboration between X3J3 and the non-US members of WG5, and we have decided, initially as an experiment, to have one joint meeting of WG5 and X3J3 each year, in February, in addition to WG5's normal summer meeting. This joint meeting will be a full WG5 meeting AND a full X3J3 meeting. Part of the meeting will be held under WG5 rules, with me in the chair, and part will be held under X3J3 rules with Jerry Wagener in the chair. The exact arrangements will need further discussions between Jerry and me, but we very much hope that this will enable members of WG5 and X3J3 who might not otherwise meet each other to work together on various technical issues in a manner which will help subsequent electronic technical work to progress more easily. More about this later. Before moving on to technical matters I must make mention of the excellent arrangements in Dresden. Our local host, Wolfgang Walter, ensured that we remained active to the end by making us walk for 20-30 minutes (slightly up hill!) each morning to get to the meeting place, by providing generous refreshments throughout the day, and champagne (yes - really!) to encourage us on the walk (downhill) to our hotel at the end of the day. He also organised a spectacular excursion on Wednesday afternoon and evening, as well as an extra excursion and dinner on Friday evening for those of us who were still in Dresden. In between we got a lot of important work done! The major items of business were the three Technical Reports, the three other technical sub-projects identified in San Diego, and the determination of the requirements for Fortran 2000. In order to allow time for reflection, the six technical issues were alternated throughout the week with Fortran 2000 requirements analysis, with the latter being discussed in subgroups during most of Tuesday, in both plenary and (different) subgroups on Thursday, and in plenary on Friday, while the six technical issues were discussed in plenary and subgroup on Monday and the first part of Tuesday morning, and again on Wednesday. The TR on Floating-Point Exception Handling (N1195r1) was presented by Keith Bierman in the absence of John Reid who seemed to think that his daughter's wedding was more important than attending a WG5 meeting (;-). (He was quite right, of course.) Subject to some minor changes it was agreed that this TR was now ready to be forwarded to SC22 for PDTR balloting (equivalent to a CD ballot). The TR on Enhanced Data Type Facilities (N1196) was also approved for forwarding for PDTR balloting, subject to the incorporation of the following informative paragraph about optimization at the end of section 3.4: Note that this definition of assignment facilitates certain optimizations when the allocatable array component of the expression is allocated. In particular, (1) if the corresponding component of the variable is allocated with the same (or larger) size, its storage can be re-used without the overhead of an additional allocation and deallocation; (2) if the expression is a function reference, the processor may simply copy the descriptor instead of the allocatable array contents and omit the deallocation of this component. The TR on Interoperability Between Fortran and C (N1178) is not yet ready for submission to SC22, not least because the change of editor when Jamie Shiers was unable to continue his involvement with Fortran standardization activities resulted in work starting several months later than was the case with the other TRs. However, WG5 investigated the approach recommended as well as that proposed by HPF, and established a compromise position. It also requested the editor, Michael Hennecke, to submit a final draft based on this approach to the Convenor by the end of 1996. In accordance with our normal practice I intend to carry out an informal electronic ballot amongst the WG5 membership on each of these documents before submitting them to the SC22 Secretariat for PDTR processing. The ballots for the first two TRs will start as soon as the revised documents are available and, hopefully, will end in time for the draft TRs to be handed to the Secretariat during the SC22 Plenary at the end of September. The ballot for the third (interoperability) TR will take place in January/February next year in the expectation that PDTR processing start in March. As in the case of standards, there are two ballots required - a PDTR ballot of SC22 members, and a subsequent DTR ballot at the JTC1 level. If no major issues are identified during balloting I would expect the first two TRs to be published towards the end of next year, and the third by Easter 1998. There were three other sub-projects initiated in San Diego (parameterized derived types, conditional compilation, and interval arithmetic) and a resolution was made on future plans in each of these areas. The parameterized derived types project was established when it was decided to remove this topic from the TR on Enhanced Data Type Facilities. Steve Morgan had worked closely with X3J3 on this proposal (N1193); however WG5 decided that, although it was happy with the technical content of the proposal, it did not believe that it was appropriate for processing as a Technical Report. We therefore identified this facility, as described in an updated version of the proposal (N1207), as a firm requirement for Fortran 2000. Conditional compilation was approved in principle in Tokyo, and in San Diego David Epstein was appointed as interim editor and requested to bring two alternate proposals to Dresden. The first of these (N1192) uses a "Fortran-like" syntax, while the second (N1208) is similar to Sun's fpp (c.f. cpp). David Epstein presented the first approach, while Keith Bierman presented the second. After discussion we decided to follow the first approach, and subsequently confirmed that it is our intention that this should be an (optional) Part 3 of the Standard, rather than being incorporated within the base Standard (Part 1). Accordingly, I shall conduct an informal electronic ballot amongst the WG5 membership, and will then submit the document, modified as necessary, to the SC22 Secretariat for CD processing. As with the TRs, hopefully, I shall be able to take the draft CD with me to the SC22 Plenary at the end of September. The final sub-project established in San Diego concerned facilities for Interval Arithmetic, with Baker Kearfott as interim editor. The subgroup had produced a proposal (N1209) for this, and after some discussion we agreed that this was not appropriate for either a Type 2 TR or an (optional) fourth part of the standard, but should be a firm requirement for Fortran 2000. Which brings us to Fortran 2000. In San Diego the high number of CD comments (597!) meant that there was insufficient time to properly consider this issue, and all that was done was to undertake a somewhat crude vote which, nevertheless, enabled a number of high priority items to be identified. In Dresden we had a substantial number of additional requirements, and we began by splitting into four subgroups (carefully arranged to split national delegations across as many subgroups as possible), each of which was charged with identifying a theme, or themes, for the Fortran 2000 revision, and a maximum of fifteen requirements from the total requirements database that best reflected the chosen theme(s). As might be expected, there was a considerable difference in the themes and resulting requirements when we briefly reconvened on Tuesday evening, but subsequent analysis showed that there was less difference than first appeared. In particular, although expressed very differently, the aims for Fortran 2000 expressed by the subgroups could all be (loosely!) encompassed by the following two themes, which the meeting subsequently endorsed as the guiding principles for Fortran 2000: Fortran 2000 should be - A language for high performance numerical, scientific and engineering programming - A modern language with high quality data abstraction and user extensibility features What was more interesting was that all of those items on the requirements listed by at least three of the subgroups had also been identified as very high priority items in San Diego. Armed with this information, we then spent most of Thursday afternoon in three subgroups, one for each of the two themes already identified and one for other topics which did not neatly fit into one of these areas. The reports produced by these subgroups were then amalgamated (by me) into a first draft requirements specification. When we reconvened on Friday morning it was generally agreed that the resulting document was too long and lacking in focus, and so the subgroup reports were removed, and the resulting document (N1215) will be forwarded to X3J3 as our initial requirements document for Fortran 2000. It identifies eight firm requirements for Fortran 2000: Floating-Point Exception Handling (as in the TR) Enhanced Derived Type Facilities (as in the TR) Interoperability with C (as in the TR) Derived type I/O Asynchronous I/O Procedure variables / pointers to procedures Interval arithmetic (initial proposal as in N1209) Parameterized derived types (initial proposal as in N1217) We also decided to establish three subgroups on the same topics as the Dresden ones which will be charged with "producing additional requirements, including, where appropriate, the production of detailed proposals and/or specifications, in order to better meet the broad objectives of Fortran 2000, as indicated above. The subgroup editors are as follows: Numerical issues: Baker Kearfott (USA) Data abstraction: Manuella Zuern (Germany) Miscellaneous: Christian Weber (Germany) Kurt Hirchert has offered to establish email reflectors for these subgroups, and an initial membership of each was established during the meeting. As soon as the reflectors are set up a call for additional members will be issued. The subgroups will provide final recommendations by the beginning of January in order that these can be circulated to WG5 (including, of course, all X3J3 members) in ample time for the joint meeting of WG5 and X3J3 in February. At that meeting, WG5 will make its final definition of the requirements for Fortran 2000, including any necessary alterations to the schedule. Finally, a word about future meetings. As already indicated, we have decided to hold joint meetings with X3J3 in February (they normally meet in February, May, August and November) in addition to our regular meetings in the summer. Gerhard Schmitt (Austria) has kindly offered to host a meeting in Vienna next year, while Lars Mossberg (Sweden) has offered to host the summer 1998 meeting. Subject to confirmation of exact dates for the last three, we therefore decided to hold our meetings during the next two years as follows: 1997, February 10-14 Las Vegas, NV (joint with X3J3) 1997, July 21-25 Vienna, Austria (dates subject to confirmation) 1998, February USA (provisional, joint with X3J3) 1998, June/August Trollhattan, Sweden (provisional) ---------------------------------------------------------------------- Dr Miles Ellis Director: Educational Technology Resources Centre University of Oxford, 37-41 Wellington Square, Oxford OX1 2JF, ENGLAND Telephone: +44 1865 270528 Fax: +44 1865 270527 Email: Miles.Ellis@etrc.ox.ac.uk WWW: http://www.etrc.ox.ac.uk/Personal/Miles/Miles_Ellis.html