Following AT&T’s decision to adopt its SPARC microprocessor as its next generation high-end hardware platform, Sun Microsystems is widely seen as having assumed the leadership mantle from the phone company in the Unix standards world. Next up will be the Applica tions Binary Inter face… John Abbott tells the story. Sun’s policy of ‘throwing technology […]
Following AT&T’s decision to adopt its SPARC microprocessor as its next generation high-end hardware platform, Sun Microsystems is widely seen as having assumed the leadership mantle from the phone company in the Unix standards world. Next up will be the Applica tions Binary Inter face… John Abbott tells the story.
Sun’s policy of ‘throwing technology over the wall’ just makes analysts want to go out and sell all our stock, said Scott McNealy, president of Sun Microsystems in his keynote speech at Unix Expo last month. And at the moment at least he was right Sun’s recent technology development plans with AT&T were announced on the same day that the disastrous stock market crash, first hit the news, and didn’t prevent its shares plunging from $40.50 to $34.25 that first week: nor did the further news a week later that the giant Xerox Corp had added support for Sun’s SPARC chip technology – Sun’s share price dropped even further to $26.75 even as McNealy was speaking. The Mountain View, California company, however, has at least until recently been one of the best bets on Wall Street. It has grown rapidly since its inception in 1982, and this year reported sales of $537.5m, a 156% hike on the 1986 figure of $20.1m. Over the wall Yet it consistently spends years and dollars generating new technology such as the Network File System, its NeWS Windowing System, and now the Scalable Processor Architecture, SPARC, RISC microprocessor, only to throw it over the wall by allowing other vendors, including its competitors, to license it. Clearly, Sun benefited more from widespread dissemination of NFS than it would have done by keeping it as a proprietary advantage for its own customers only. By opening up the technology it created the initial momentum for the establishment of a de-facto standard for distributed systems, and thereby broadened the market. Why, however, should it let other manufacturers use its SPARC microprocessor, reducing any performance advantages it might have kept to its own product? The first reason is practical: Sun cannot go full-scale into the microprocessor business on its own, so it must licence others (currently Fujitsu, Cypress and BIT) to produce the chip, ensuring enough market opportunity to make it worth their while. The second reason is the software associated with the chip. By interesting AT&T in the SPARC processor (a real coup considering AT&T’s own Western Electric division was carrying out its own RISC development, now discontinued), Sun has placed itself right in the centre of future developments in the Unix operating system as well as getting some of its own SunOS features (such as NFS and X.11/NeWS) and ABI incorporated into the AT&T Unix System V Interface Definition. A straw poll at Unix Expo revealed that most people perceived Sun as taking the major lead in the future development of Unix, over and above even AT&T itself. The joint development effort with AT&T involves a three phase plan of action. In the first phase, Sun will release its long awaited merge of System V and Berkeley 4.2 (and SunOS features) for the SPARC processor. Work on this is apparently all but complete, and the product should be shipped by the second quarter of 1988. Phase two will see AT&T’s incorporation of the new features into a unified version of Unix System V in the first half on 1989. Phase three is now being worked on by Bill Joy and a Bay Ana development team, and involves the development of a totally distributed Unix environment that allows for multiprocessing according to McNealy, and results will also be seen in 1990. The unified Unix versions will allow the implementation of another important event in the Sun/AT&T effort: Applications Binary Interface, ABI, described by McNealy as a typically out-of-control Sun manoeuvre. ABI aims to enable Unix software conforming to the Interface to be fully binary compatible across machines, just as MS-DOS software is on Intel processors. ABI however goes beyond this, embracing C language and library definitions, operating system interface, file locations and
formats, and networking, windowing and graphics facilities, which will enable developers to create applications with a familiar look and feel but strong common characteristics – something that could make Unix machines like an open Macintosh. The engineering effort, to be taken on by 100 of AT&T’s 600 Unix programmers, according to Computer Systems News, is to pull all the technology, which mostly exists already, together. MS-DOS vendors Which application interface component will be used has not yet been decided but will be announced during the first quarter of next year at a series of seminars for developers to be held by Sun. Massachusetts Institute of Technology is working on such an applications interface toolkit based on X Windows. McNealy said that ABI had attracted interest from MS-DOS software vendors who have been looking towards Unix for some time, and written (or re written) software in C. With binary compatiblity, distributors could become excited about the retail opportunities for the first time, especially with parallel efforts for an 80386 binary standard from Microsoft and Interactive Systems Corp working on the merged Unix/Xenix, 386 binary and ABI binary standards would be drawn together by Phase 3, said McNealy, who also predicted that computer manufacturers would be encouraged to produce small Unix machines once a binary standard had become established. The SPARC processor, initially at least the hardware base for these efforts, has has been boosted by support from not only AT&T and Xerox, but also Arete Systems. Fujitsu, the major source of chips at the moment says it plans to be shipping 15,000 a month by the end of 1988. Multiple licencees for the chip is an important selling point for companies such as Arete which will buy from Fujitsu and Cypress: it means that they do not have to rely on one supplier to keep up with technology – custom CMOS, ECL, and Gallium Arsenide implementations could take the SPARC MIPS rating up to three figures – and also allows for some degree of competition.