From FS300022@Sol.YorkU.CA Thu Oct 21 21:05:00 1993 Received: from sol.yorku.ca by dkuug.dk with SMTP id AA25716 (5.65c8/IDA-1.4.4j for ); Fri, 22 Oct 1993 06:05:21 +0100 Received: from Sol.YorkU.CA by Sol.YorkU.CA (PMDF #12410) id <01H4E80KSRFK8ZEFWC@Sol.YorkU.CA>; Fri, 22 Oct 1993 01:05 EDT Date: Fri, 22 Oct 1993 01:05 EDT From: FS300022@Sol.YorkU.CA Subject: Just Honour for Fortran Developer, John Backus - Lest We Forget To: SC22WG5@dkuug.dk, martin@ocfmail.ocf.llnl.gov, bbuckley@agb.RoyalRoads.ca, waiello@vega.math.ualberta.ca, aubryp@pbs.dfo.ca, jthimer@iras.ucalgary.ca, ranga@enel.ucalgary.ca, melanieu@vnet.ibm.com, GWARREN@torolab2.vnet.ibm.com, supy@nanonic.ucs.ualberta.ca, dtm@castle.ed.ac.uk Message-Id: <01H4E80KSRFK8ZEFWC@Sol.YorkU.CA> X-Envelope-To: SC22WG5@dkuug.dk X-Vms-To: @X3J3-WG5.LST X-Vms-Cc: FS300022 X-Charset: ASCII X-Char-Esc: 29 It is customary when a well-known professional gets a high honour, for all concerned to applaud; from a personal perspective nobody (including me) would object. Yet from a scientific and technological viewpoint, the development of FORTRAN by John Backus (circa 1955-1960) and more especially its aggressive marketing by IBM, was an unfortunate, retrograde step. Concurrently with that development an International consortium of Mathematicians and Engineers produced the specifications of what became ALGOL-60, a language which is intrinsically superior (from the fundamental standpoint of the computational requirements of Turing Machine processes) not only to the Fortran I & II of the 1960's, but also superior to FORTRAN-90. This latest version does, finally (after 30 years of being constrained by static memory allocation) have some form of Dynamic Memory Allocation (a fundamental requirement of Turing Machine processes as distinct from Finite Automaton processes), but it is implemented clumsily compared with the generality and elegance of ALGOL's Nested Block structure. For a few years the ACM would only accept programs/algorithms written in ALGOL-60, but in response to commercial (anti-academic) pressure they also accepted Fortran programs. The work of X3J3 and WG5 over the past 35 years has essentially been to try to patch up the legacy of John Backus's intrinsically bad design (implied static memory allocation, multiple declaration of variables, storage association, no nested environments, etc). The battle between enlightened academic judgments and commercial compatibility interests has been (and still is) a recurring undercurrent of X3J3/WG5 meetings; e.g. the battle over obsolescent/deleted features within the broader context of what was (euphemistically and myopically) called "Language Evolution"; this battle has being going on for several years and is continuing today. It is fair to say that the emergence of FORTRAN as the most popular scientific/engineering programming language was the result of IBM's stranglehold on the marketplace that began in the 1960's (coupled with the "Design Flaw in the IBM/360" - Static Relocatability - Saul Rosen, Comm.A.C.M., 1975). This detrimental influence has retarded the natural and orderly growth of the Art and Science of computing for what is now 35 years, and by current trends is destined to soon become at least half a century. When the X3J3 and WG5 members (and the many people that they represent and influence) emerge from their myopic addiction to FORTRAN, when they step back to ALGOL-60, SIMULA-67 and ALGOL-68, then (and only then) will the Art and Science of Computing really begin to take significant steps forward, for notwithstanding the advances in hardware technology, the design of most current systems is still rooted in the static relocatability, static memory allocation regimen of the 1960's. As Edsger Dijkstra said about 10 years ago: "Microcomputers are Not Great". In contrast to the commercially popular computers, the Burroughs machines (B5000,B6700,B7700,A-series) throughout the past 35 years have remained intrinsically better designed than almost all other machines. Anybody who has worked on a Burroughs mainframe will tell you of the sense of calm mastery of Error-Free computing that working on a well-designed system can impart - in sharp constrast to the neurotic, frenetic, error-prone sense of "will it ever be right" that characterises working with currently popular systems (including FORTRAN-90), and that keeps the X3J3/WG5 circus touring the world. As for UNIX & C; well C has only a 2-level (Global/Local) environment structure (less general than ALGOL-60's multiply nested environments), and UNIX (in common with MS-DOS) has the fundamental design flaw of copying the parent's environment when a daughter process is created, thus preventing communication between daughter and parent via the latter's environment. When the IBM Chairman, Louis V. Gerstner Jr., exploits John Backus's prize to claim that: " . . . it helped to establish a tradition of technological leadership at IBM -- a tradition that carries on to this day.", we are hearing a piece of commercially motivated lying. IBM was the last company to have Virtual Memory (because of the Design Flaw in the IBM/360), and for the same reason they were last to introduce interactive computing (formerly "timesharing"). Their current dominance of the PC marketplace has inflicted the poor design of the INTEL 8080 upon millions of users (the 640 Kilobyte memory limit, etc), while the small, elite populace of MAC users (i.e. the Motorola 68000 architecture) experience the advantages of an intrinsically superior design (as did users of Burroughs machines in the mainframe era). Commercialism Corrupts Human Intelligence, Geoffrey Hunter (formerly of the Atlas Computing Laboratory, Manchester, England), Department of Chemistry, York University, Toronto, Ontario, Canada M3J 1P3. EMAIL: FS300022@YUSOL.YORKU.CA Tel:905-736-5306