To: WG5 From: Miles Ellis Subject: Development of Fortran beyond Fortran 95 WG5's current plans are to finalise the draft CD for Fortran 95, incorporating the items included in Corrigenda 1 and 2 (as well as those which we may submit for Corrigendum 3) together with a number of relatively small extensions to the language - and a few deletions. The current plans are that this should be followed by a major revision should be in 2000/2001 (Fortran 2000). However, in the last few months it has become increasingly clear that there are several major features that really cannot wait until 2002/2003 _ allowing the usual 2-3 year delay between the publication of the Standard and the availablility of a range of compilers conforming to that Standard. One school of thought has been arguing that we should delay Fortran 95 and add these features (e.g. exception handling, various object oriented features, etc, etc) before it is sent out for a CD ballot. However, the majority view is that we must stick to our widely announced schedule. Nevertheless, there is an urgent need to address these issues. One problem, as I have hinted above, is that vendors have, in the past, been very reluctant to make significant modifications to their compilers until the exact syntax and semantics of new features are, as it were, cast in stone (and given the problems that we had in the late 1980s with Fortran, who can blame them!). However, this does mean that new, and often quite complex, features are being incorporated into the Standard without any real experience of their use, with the inevitable result that bugs and holes are discovered when actual implementations are being prepared and when they are in use. The present process seems very like producing a new, and major, software product without even an alpha-test version, never mind a beta-test! It is hardly surprising that it has produced a substantial number of minor bugs, and a smaller number of more serious ones. My proposal is to use the tools at our disposal rather better. Specifically, I propose that WG5 should establish two or three small task-forces, each charged with producing a Technical Report (type 2) on one of these major new features. It would be WG5's declared intention that the syntax and semantics described in such a report would be incorporated in the next revision of the International Standard for Fortran, unless experience in the implementation and use of the features covered by the TR indicated that a change was required. Such an approach would enable major new features to be developed more quickly than at present, and would allow for "alpha" and "beta" testing before they were finally incorporated in the Standard, while also providing a reasonable guarantee that implementations of the features described in the TR(s) by vendors would not be wasted effort, as the features would not be subsequently changed unless the actual experience showed that this was necessary. Informal discussions with several vendors at the most recent meeting of X3J3 indicated that they might welcome such a procedure, as they are well aware of the user demand for some of these features, but feel unable to meet it until it has been officially standardised. The acceptability of this approach to future work has been confirmed, in principle, by the SC22 Secretariat, subject to the proviso that the the concept should be proposed to SC22 at its Plenary in September 1995, together with any necessary New Work Proposals _ one for each Technical Report planned.