From jwagener@ionet.net Mon Dec 30 22:41:51 1996 Received: from mail.ionet.net (ultra2.ionet.net [206.41.128.42]) by dkuug.dk (8.6.12/8.6.12) with ESMTP id WAA16245 for ; Mon, 30 Dec 1996 22:41:42 +0100 Received: from Zvyvogs (tsip164.ionet.net [206.28.164.173]) by mail.ionet.net (8.8.4/8.7.3) with SMTP id PAA25073 for ; Mon, 30 Dec 1996 15:41:03 -0600 (CST) X-Mailer: InterCon tcpCONNECT4 4.0 (Macintosh) MIME-Version: 1.0 Message-Id: <9612301548.AA14206@Zvyvogs> Date: Mon, 30 Dec 1996 15:48:14 -0600 From: "Jerrold L. Wagener" To: sc22wg5@dkuug.dk Subject: Van Snyder's wish list Content-Type: Multipart/Mixed;boundary=part_AEED94BE000AE1DD00000002 --part_AEED94BE000AE1DD00000002 Content-Type: Text/Plain; charset=US-ASCII Content-Disposition: Inline Happy New Year to all! To start the new year here is a (hopefully) provocative note. I have recently heard from Van Snyder (Nasa/JPL), arguably the reviewer with the most comprehensive comments during the public review of Fortran 90. Perhaps exhausted from his labors during that review, he seemed to drop out of sight (at least from mine). But now he's back, and Van - welcome. Van is a firm friend of Fortran, though that might not be obvious at first glance from his sometimes-acerbic attacks on certain aspects of (then draft) F90. But then, Van was ahead of the times (he may still be ahead of the times). Van has given me permission to announce to WG5/X3J3 his web page entitled "Wish list for Fortran 2000": http://gyre.jpl.nasa.gov/~vsnyder/ fortran. (He has recently announced it on comp.lang.fortran, but I had a difficult time accessing it - hopefully those bugs are now exterminated.) This web page should be required reading for members of WG5 and X3J3, especially those who will be attending the Feb joint meeting. Though he can be harsh with us (the standards community) at times, he is a firm friend of Fortran as a user who cares about the evolution of the language and has ideas about that evolution that are provocative yet practical (and did I mention often ahead of the times - he saw the data abstraction potentials for Fortran, and how to do it right, before most of the rest of us and before object-orientation and C++ took over the world). His web page is a well- organized compendium of his current ideas/suggestions for Fortran. The page starts with a index of 22 items, each of which can be followed, some only one level and others more completely. While I certainly do not mean to imply that I like everything (or even anything) there (actually I do like a couple of things :), they are virtually all worth reading. Let me briefly comment on some of the ones I found most interesting/provocative (in the order they appear in his list): - gratuitous constraints Van laments the language complexity due to unnecessary constraints (some of which may have been justified decades ago). - object-oriented Fortran Van's counsel is to follow the Ada model, not the C++ model. - threads Van asks the question "why limit async to I/O?" Why indeed? He gives an interesting suggestion for generalizing async I/O. - is FPP or CoCo necessary? Van doesn't question the requirement, but the implementation. He suggests it can be made much simpler still. - Fortran could have referential invariance "Referential invariance" is the term he now uses for the thing for which he took us most to task in his F90 review. (IMHO he was right then, but we didn't listen; IMHO he's still right. Did I mention that he's sometimes ahead of the times?) It would appear that "it's too late" for his approach, but Van says maybe not. - subtypes of integer need regularization Van says getting unsigned integer is simple. (Can other types be far behind?) So much for brief comments on six of the 22 - comments intended to make you curious as to what is really in there. Unfortunately the "closing" of the requirements process means that some of these probably can't be considered (emphasize *considered*, not intended here to advocate *adopted*) for Fortran 2000. But a number impinge on likely requirements for Fortran 2000 from the requirements process, and Van may have some useful insight (outsight?) for us on those. Jerry --part_AEED94BE000AE1DD00000002--