A comfortable assumption about Japan Inc widespread in the US and Europe is that while it leads the world in hardware design and low-cost production, when it comes to software, Japan lacks the basic skills – even the national temperament – to master the technologies. The consequent argument has been that since hardware prices are falling so fast, an ever-increasing proportion of systems costs are accounted for by software, so that it really doesn’t matter if Japan dominates the world microelectronics industry, because in the limiting case, systems cost will be represented almost entirely by software, and Japan will be left with 100% of nothing. The comfortable assumptions promise to be blown apart by another of those Japanese national efforts that make Britain’s Alvey programme, Europe’s Esprit and even America’s Microelectronic and Computer Technology Corp look like bad jokes perpetrated by rank amateurs. Japan’s Sigma project is as significant as the effort inspired by the Ministry of International Trade & Industry in the early 1970s to rocket Japan to the number one spot in the high-volume semiconductor market, and this time the target is the world software industry. Japanese interest in software development methodologies is heightened by the fact that users tend to demand custom applications and are averse to using standard packages – even Japanese ones, let alone imported ones.
There is therefore an enormous home market incentive behind the Sigma – Software Industrialised Generator and Maintenance Aids project – once again sponsored by MITI, to develop a methodology for much cheaper production of software. As in the US and Europe, the applications back-log is growing, and the rate of growth is rising as Japanese banks and finance houses look to find international homes for the massive weight of savings by Japanese individuals. There are even local cultural pressures for a low-cost software development methodology – technical specialists are regarded well below generalists working for big companies, and higher salaries for technical people – around $26,500 for programmer – are insufficient to attract the high-flyers, leading to a forecast 600,000 shortfall in development staff by 1990. MITI has therefore committed $165m over five years to the Sigma project develop a shared software development system, and some 150 companies have invested in the project, including all the Japanese hardware manufacturers, and even some foreign manufacturers. The project started in October 1985, with a very substantial tripartite infrastructure – the Sigma centre, the Sigma user site and the Sigma network. The Sigma centre is composed of five subsystems, each running on a mainframe provided by one of the hardware vendors: the database subsystem runs on a Hitachi M-260D mainframe, the network subsystem is supported by a Nippon Telegraph & Telephone DIPS/CIP-2, the development environment subsystem runs on a Fujitsu M-360R mainframe, an NEC Acos 630-10 is being used for the accounting and statistics subsystem, and Fujitsu has also provided a demonstration subsystem. Hardware also includes a DEC VAX-11/785 and a Data General MV/10000 SX. Thus equipped, the Sigma centre is due to begin operations next next month.
A standard Sigma workstation has already been specified to run the evolving Sigma operating system, which is based on Unix System V.2 with Berkeley 4.2BSD extensions; 50 prototype workstations to the standard Sigma specification are going out from six manufacturers – Fujitsu, NEC, Sumitomo Electric, Toshiba, Omron Tateisi and Mitsubishi Electric – this month. The Sigma network is the communications link which ties all the elements together. As well as creating standard software development tools for project and resources management, automatic documentation generation and so forth, Sigma will provide a database service listing available hardware and software and reference material on Japanese and international networking, programming and operating system standards, and a full range of network service functi
ons including file transfer, support for remote users, and electronic mail services. The Sigma operating system is currently at release 0, but already adds Japanese language processing to System V.2 and BSD4.2. Extensions still to be developed – or adopted from abroad – include multi-media windowing, a virtual printer interface, the System V.3 and 4.3BSD distributed file extensions, a Japanese language C compiler, and the seven-layer Open Systems Interconnection standard. Communications to be supported include dial-up and leased point-to-point links, X25 packet switching, IEEE 802.3 local area networking and the TCP/IP transfer protocol. The GKS Graphics Kernel System will be supported, and graphics functions will be provided for creating flow charts, logic diagrams and so forth on-screen. The hardware pre-requisites for the Sigma workstation prototypes were extremely comprehensively defined. It had to have a processor with internal 32-bit registers, support for floating point arithmetic to the IEEE standard; at least 4Mb main memory and a logical address space of at least 8Mb, of which at least 6Mb had to be available to the user; a Winchester with at least 20Mb available to the user; a floppy disk drive capable of reading and writing IBM 8 floppy disk format; support for some kind of tape back-up – exchangable cartridge disk, open reel magnetic tape, or quarter or half-inch streamer; a display with at least 1,024 by 768 pixels and character capacity of 40 24 by 24 pixel characters by 24 lines, with support for windows and at least 16 colours as mandatory options; Japanese Institute of Standards keyboard with at least 10 function keys; mouse with two or more buttons; two or more serial RS-232C interfaces, at least one of them capable of supporting a modem at 300 to 9,600 bps full-duplex, and a parallel Centronics interface; a local area network interface to the IEEE 802.3 standard; a DDX-P circuit switched X21bis or X21 interface; an optional GP-IB interface of at least 10Kbps; and – a very Japanese touch – an external security device such as identity card reader to prevent unauthorised use of the software. The prototype workstations will be used for development of tools for networking and documentation support, and for internetworking verification with the stations from different vendors connected together. Sigma tools fall into two classifications – common tools, for networking, documentation support, project management, library management and a man-machine interface library, and a host of application tools – languages and so forth.
Development of each class of tool has been contracted to a different software house. Major companies were awarded the Common Tool contracts – Oki and Nomura Computer Systems for network tools, Fuji Xerox for documentation support tools; Chiyoda Chemical Engineering & Construction for project management tools. All these tools will be available on-line to all Sigma users from every Sigma workstation. Application tools include Cobol generators, editors and analysis tools, documentation, JCL generators; scientific and technical calculation tools, including support of Fortran source libraries; process control application tools; and personal computer tools. All the tools have been developed in C – currently the portable C that comes with Unix System V, but this will be converted to any international standard which which emerges. Conceptually therefore, the Sigma project is by no means dissimilar to much of the UK’s Alvey project. The clear difference is the extent to which all the major Japanese companies are involved, and the degree of central control that ensures that everyone involved knows precisely what everyone else is doing, and exactly what common goal is in view. By comparison, Alvey looks hopelessly fragmented – and in contrast to Alvey, where outsiders regard it as a good thing in principle, but question whether it will really achieve anything, Japan is thoroughly committed to the Sigma project, and very definitely believes that it will make a major contribution to solving