UNIX STANDARDS IN THE 1990s
February, 1992: Unix4.doc (Paul)
February 16-20, 1992: Ustd1.doc (Will)
March 28, 1992: Ustd3.doc (Will)
December 1994: ustd5 (minor revision)
The latest International Computerquest Corporation analysis of the market for Unix-based computers landed on three desks on the same morning. Noel Sharp, founder, chief executive officer, chief engineer, and chief bottle washer for the Superbly Quick Architecture Workstation Company (SQUAWC) in Mountain View, California hoped to see strong growth predicted for the market for systems designed to help architects improve their designs. In New York, Bo Thomas, senior strategist for the Unix systems division of A Big Computer Company (ABC), hoped that general commercial markets for Unix-based computer systems would show strong growth, but feared that the company's traditional mainframe and mini-computer sales would suffer as a result. Airborne in the middle of the Atlantic, Jean-Helmut Morini-Stokes, senior engineer for the Unix division of European Electronic National Industry (EENI), immediately looked to see if European companies would finally have an impact on the American market for Unix-based systems. After looking for analysis concerning their own companies, all three managers checked the outlook for the alliances competing to establish a Unix operating system standard.
Although their companies were alike only in being fictional, the three managers faced the same product standards issues. How could they hasten the adoption of a Unix standard? The market simply would not grow until computer buyers and application software developers could count on operating system stability. And how could their companies benefit once the standard was determined? The following report summarises issues in the Unix-based computer market during 1991 and early 1992.
1. UNIX BACKGROUND
1.1 What is Unix And What Is It Used For?
Unix is a combination of a computer operating system and utility programs. Written mainly in a technical computer programming language called "C", Unix programs can be moved among different computers, thereby facilitating the work of software developers and end users. Unix allows a single user to operate several computing jobs at the same time, which is known as multitasking.
Since its commercial appearance in 1975, all major hardware vendors have introduced products that operate under Unix. Versions of Unix exist for all types of computers, including mainframe, mini, and microcomputers; almost all technical workstations; and most supercomputers. By 1988 there were more than 600,000 Unix installations worldwide and by 1991, the list of Unix commercial users included most Fortune 100 companies. Despite this level of success, Unix has still had a relatively minor impact in commercial markets by 1991 and had yet to live up to its commercial promise. Unix was still the operating system of the future, not the present.
1.2 Unix History
The Unix operating system was created in the American Telephone and Telegraph (AT&T) Bell Labs in 1969 by Ken Thomson, Dennis Ritchie, and their colleagues, who aimed to develop an operating system that would isolate computer hardware from application software. The first edition of the system was documented in a manual in 1971 and the sixth edition became the first commercially available version of Unix, in 1975.
AT&T at first viewed Unix as an academic and research tool. Lacking major commercial goals for the system, AT&T licensed Unix widely for a nominal fee. AT&T also was inhibited from entering commercial computer markets because of agreements that it had reached with the Federal Trade Commission (FTC) concerning antitrust issues in its telecommunications business. Licensees paid for the right to use the Unix "kernel" (the core operating system) and then developed enhancements to the program, such as new utilities, different algorithms, security modifications, real-time capabilities, and symmetric multiprocessing. The first version of Unix to gain favour in academic institutions and other research-oriented sites emerged from the University of California at Berkeley, which developed and distributed BSD (Berkeley Software Distribution) Unix.
Contrary to AT&T's expectations, Unix also became popular in commercial settings. The first entrant was Xenix, a version of Unix designed to operate on 16-bit microcomputers that was co-developed by Microsoft and the Santa Cruz Operation (SCO) and introduced in late 1980. AT&T also introduced Unix System III in 1980, while the first version of Unix designed to operate on mainframe computers was released in early 1981. In 1984, AT&T introduced Unix System V Release 2 (SVR2), marking its major entry to the commercial market as well as signalling its intent to promote System V as the product standard. In 1985, AT&T followed with System V Release 3 (SVR3). In addition, AT&T designed the System V Interface Definition (SVID) and published the System V Verification Suite (SVVS) for product testing, obliging developers to pass SVVS before they can begin shipping any derivative using the SVR3 licence. AT&T could force licensees that did not comply with the testing procedures to withdraw their product from the market. Following relaxation of the FTC-imposed restrictions on its participation in the computer market, AT&T introduced a microcomputer designed to use Unix about 1986, competing with IBM-compatible and Apple microcomputers that used the DOS and Apple operating systems.
Despite its lead, AT&T could not dominate the Unix market. With the widely available Unix kernel, many other hardware manufacturers and software specialists introduced Unix variants that were not derived from System V and so were free of SVVS restrictions. By 1991, more than 250 versions of Unix had been introduced for commercial or research use.
Many factors contribute to Unix's attraction. Companies can take advantage of open systems architecture, which is Unix's ability to operate on different hardware platforms, thereby avoiding commitment to a single hardware vendor. Consequently, a company can choose hardware and software based more on price-performance ratios and less on the compatibility of current and future applications. In addition, Unix can reduce the costs of personnel training and data information management by 50% or more. These costs constitute a major expense, sometimes exceeding hardware and software costs. Software developers can use Unix to reduce development costs, because it potentially enables them to write one application for many systems rather than modify it for each hardware platform. Nonetheless, several serious obstacles confronted Unix commercial acceptance in 1991. The following tables summarise the advantages and disadvantages of Unix.
Table 1. Factors Contributing To Unix
1. Open Systems. Unix runs on almost all vendors' computing platforms.
2. Portability. The same program runs on many different computers.
3. Functionality. Unix supports multitasking, multiuser applications, networks, and high level applications. All were once only available with minis and mainframes.
4. Government Acceptance. The US Government and the EC support Unix.
5. RISC. RISC-based workstations, which generally use a version of Unix, are becoming popular as their prices drop, offering a better price/performance ratio than other systems.
6. User Interface. Several user-friendly interfaces have been developed.
7. Applications. Many popular applications have been released for Unix systems. In addition, many Unix configurations accommodate the use of DOS.
8. OS/2 and Windows. Unix major competitors are in their infancy.
Table 2. Obstacles To Unix Commercial
1. Many Variants. Many existing variants are not totally compatible.
2. Applications. Despite growing availability, less software is available than for other operating systems.
3. Employee Resistance. Corporate personnel is not familiar with Unix.
4. Installed Base. Billions of dollars invested in other systems.
5. Security. Unix has acquired the reputation of being weak in the area of security.
xx 6. Technical Limits. As a relatively new system, Unix has technical problems to be overcome.
1.3 Unix Standards Issues
Two related issues provided the chief obstacles to Unix's success: lack of product standards and lack of software availability. The existence of many versions hindered Unix acceptability in business applications. More than 30 variants of Unix were available commercially in 1991, with versions offered by virtually every major hardware vendor and many software specialists. Unfortunately, the versions were not totally compatible. Software applications written for one Unix variant could not operate on other Unix systems without expensive and difficult modifications. Moreover, moving applications among different hardware platforms is often unsuccessful, even in the Unix environment. The lack of a product standard produced the first major impediment to Unix acceptance, because end-users incurred major cost and problems in integrating their applications.
The small (although growing) number of applications that operate on Unix systems created the second severe impediment to full commercial adoption. In the computer industry, the most important measure of operating system success is the number and variety of application software programs that have been developed for it. Many information system managers would not adopt Unix because of the lack of available application software.
The scarcity of application software stemmed from the lack of a product standard. Software developers find it extremely costly and difficult to support multiple versions of the operating system. Software houses were slow to develop business applications designed for Unix operating systems and slow to upgrade the Unix software that they had introduced.
2. UNIX ALLIANCES
2.1 Open Software Foundation versus Unix International
The first major effort to promote Unix standardisation began in 1984, when several leading European, American, and Japanese computer manufacturers formed the X/Open group. Leading members included Siemens, AT&T, and Digital Equipment Corporation (DEC). IBM was notably absent from the founding membership, and did not join until several years later. X/Open planned to encourage the development of standard core features for commercial Unix systems, in order to increase available software and to secure portability of the applications. X/Open adopted AT&T's System V as the basis for the standard. One interpretation of this choice is that X/Open's members decided to run the risk of exploitation by AT&T rather than by IBM.
Although X/Open continues to exist, its consensus approach to standardisation began to break down in October 1987, when AT&T and the technical workstation leader Sun Microsystems announced their intention to pursue co-development of a standard Unix operating system based on AT&T's System V, Berkeley's BSD 4.2, and the graphical capabilities of Sun's Sun-OS. The new system would be available to other companies under proprietary license. AT&T and Microsoft also agreed to standardise Xenix and AT&T's microcomputer Unix. The goal was to bring the most popular versions of Unix together, although each partner had different motives. AT&T wanted allies to back up its product and attract more software developers. Sun wanted to secure the long-term success of its new SPARC (Scalable Process ARChitecture) microprocessor.
Several major vendors feared that AT&T would gain a competitive edge in software and hardware development. Many firms also feared that the joint effort would make Unix more dependent on Sun's hardware. In response, IBM, DEC, and five other hardware manufacturers established the Open Software Foundation (OSF) in May 1988. The nonprofit consortium aimed to develop a standard version of Unix that did not draw on AT&T's proprietary technology. By the end of 1988, two more major firms had joined the OSF as full sponsors, and many firms joined in secondary positions. The full sponsors would contribute $13.5 million over three years and would control the Board of Directors, while secondary memberships were available for as little as a few thousand dollars. Unlike X/Open, OSF planned to produce an operating system that it would licence to its members, rather than function only as an advisory body.
Efforts to enroll AT&T in OSF were fruitless. OSF offered to use System V as its core system, but wanted all partners sharing in future developments while AT&T wanted to control development of System V. Instead, the OSF decided to base its system on IBM's AIX, with members encouraged to submit suggestions. By 1991, OSF membership reached about 200 computer manufacturers, peripheral and component hardware manufacturers, software developers, and standards organisations.
AT&T and Sun responded by allying with several other computer vendors to form the Unix International, Incorporated (UI) trade association. UI members would advise AT&T, and would promote the development of System V. Its Board of Directors consisted of company executives, with votes based on financial contribution. Contributions were comparable to those required of OSF sponsors, ranging from $500,000 to $5,000 based on the class of membership (Principal, General, Associate, and Academic Associate). AT&T retained proprietary control of Unix and future modifications. By 1991, UI had almost two hundred members from more than 15 countries. Members included computer manufacturers, software developers, industry consultants, academic and research institutions, and government bodies.
Many less powerful companies joined both UI and the OSF. Notable examples are major microprocessor and software developers such as Motorola, Intel, and Oracle. However, there was no overlap among the full sponsors of the OSF and the principal members of UI.
The OSF and UI aimed to control the future of Unix market by creating the standard versions of the operating system and supporting software. In November 1989, AT&T introduced a commercial version of System V Release 4 (SVR4). AT&T expected SVR4 to be compatible with almost 80% of Unix installations. In late 1990, the OSF consortia released a commercial version of OSF/1, based on IBM's AIX and Carnegie Mellon's MACH operating systems. Although several attempts were made to combine UI and the OSF and coordinate development of the competing operating systems, no successful negotiations had occurred by 1991. The battle for standardisation will be the issue of 1990s.
Table 3. OSF and UI Alliance Founders
Open System Foundation (May 1988) Unix International (December 1988)
Hewlett Packard Amdahl
Apollo Control Data (CDC)
Groupe Bull NCR
Hitachi (late 1988) Olivetti
Some other notable members Some other notable
Hitachi (Japan) Santa Cruz Operation
Wang NEC (Japan)
Some firms that are members of both alliances
Data General, Motorola, Intel, Silicon Graphics, Swedish Telecom, Stratus
2.2 Desktop Computing
In 1991, several computer firms sought to emulate the experience of the Intel-based microcomputer market in guiding the development of the emerging desktop computer market. Early in its history, the Intel market adopted IBM's Personal Computer design and Microsoft's operating system as standards. As a result, multiple software and peripheral makers committed to supplying microcomputers that followed the Intel-based standards, and microcomputers based on Intel CPUs became the largest market for computers of any type because of these commitments. The technical workstation industry, by contrast, has not adopted meaningful operating system and CPU standards, and has not been as successful.
In April 1991, a group of 21 companies led by Compaq, the Digital Equipment Corporation (DEC), Microsoft, and MIPS initiated an effort to set standards for future desktop computers. Their Advanced Computing Environment (ACE) initiative includes developing a RISC-based computer standard using the MIPS CPU and the OSF's version of Unix as the operating system. ACE also includes a specification for low-end microstations that is based on Intel CPUs and Microsoft's Windows NT (New Technology) 32-bit operating system. The ACE initiative specifies that both the OSF and Windows NT operating systems will run on all computers that comply with the ACE standard whether Intel-based or RISC-based. Hence, in theory, software written for an ACE standard computer will be able to run on low-end ACE desktop computers as well as high-end ACE workstations. ACE member compliance to these standards will insure a large installed base for the alliance.
Major competitors of ACE members responded to the ACE initiative with standards-setting efforts of their own. Sun announced the development of Solaris, an operating system that would run on both Intel-based computers and on Sun's Sparc RISC-based computers. Hewlett-Packard attempted to interest ACE members in developing computers around its RISC CPU.
The strongest response came from an alliance between Apple Computers, IBM, and Motorola. Under their agreement, the companies would develop low-end computers using Motorola CISC CPUs and derivatives of Apple's Macintosh operating system. More powerful computers would use the IBM-Motorola PowerPC RISC CPU and operating systems derived from the Macintosh and from Apple's and IBM's versions of Unix. The alliance plans to replace this interim operating system with an object oriented operating system (code named "Pink") -- primarily developed by Apple -- later in the decade. Software developed for any platform would be able to run on all platforms. Therefore, like ACE and Sun's Solaris, the Apple-IBM-Motorola alliance plans to provide unified computing standards that will satisfy low-end to high-end desktop computing needs.
3. CROSS-CURRENTS OF INDUSTRY FORTUNE
3.1 Downstream: Changing Product-Market Segments
In 1991, key product segments of the computer market included mainframes, minicomputers, microcomputers (personal computers), technical workstations, and supercomputers. IBM dominated the mainframe segment, where it offered proprietary operating systems. Unix-based systems had made their inroads to the other segments.
Important product features included price (which had dropped rapidly in the recent past), computational speed (which had grown rapidly), and additional features such as random access memory (RAM) available for data processing, information storage capabilities, and graphics capabilities. Traditionally, mainframe computers such as IBM's 370 were expensive, fast, and offered large memory, computational capacity, and storage. Microcomputers were relatively inexpensive, slow, and had restricted memory and storage. Minicomputers, such as DEC's VAX systems, fit somewhere in the middle. Technical workstations, such as systems offered by Apollo and Sun, first competed with minicomputers by offering more speed, better graphics capabilities, and comparable price. Expensive supercomputers, such as systems offered by Cray, offered massive data manipulation capabilities.
The lines between the product segments are quite blurred. By 1991, technical workstations had begun to compete strongly with minicomputers, and even offered some capabilities similar to low-end mainframes and supercomputers. Microcomputers using 32-bit microprocessors, meanwhile, had acquired capabilities similar to low-end technical workstations and minicomputers. In the near future, the microcomputer and technical workstation product segments are likely to converge into the desktop computer segment, which is sometimes referred to as the "microstation" segment. The minicomputer category may drop out of future product segments.
The market for Unix-based computer systems is often segmented by noncommercial, nontechnical commercial, and technical end-uses. The noncommercial segment includes government and academic users. The nontechnical commercial segment includes data processing in the financial, medical, and other industries. The technical segment includes scientific and engineering uses in the automotive, aerospace, electronics, defence, and entertainment industries.
Table 4. Unix-Based Computer Market End-Use Segments
Noncommercial $2.1 (30%) $7.2 (40%)
Nontechnical commercial 2.1 (30%) 5.9.(33%)
Technical 3.2 (40%) 4.8 (27%)
$7.4 (100%) $17.9 (100%)
Source: Frost & Sullivan (1988)
Market segmentation can also take into account user size and intermediate buyers. Sales to large users, including corporations and public institutions, accounted for 40%-50% of Unix-based systems sales in 1987 and were expected to grow more quickly than to smaller buyers. Large users purchase hundreds of machines at one time, for whole departments or plants.
Two overlapping types of intermediate buyers also are relevant, hardware original equipment manufacturers (OEMs) and software value added resellers (VARs). OEMs are companies that incorporate a workstation vendor's hardware and (perhaps) operating system with their own hardware. Most OEMs provide system integration and technical assistance to their customers (end users). Hardware vendors give OEMs significant discounts, sometimes reaching 40% of the list price. VARs design software that tends to be dedicated to a particular line of hardware and then sell a hardware-software package to an end-user. VARs provide technical assistance to users, and often undertake system integration and installation for the users. A VAR usually specialises in certain applications such as CAD/CAM or design automation, or industries such as aerospace or architects. OEMs accounted for 40% of total shipments in 1987, while VARs accounted for about 10-20% of sales and were growing rapidly.
Major changes were taking place in product-market combinations during 1991, particularly as microcomputers and workstations found use in commercial markets. Technical workstations (including high-end microcomputers) accounted for only 5% of commercial computer purchases in the early 1990s, compared to 20% of the money spent on computing systems by engineering users, but were rapidly gaining ground in commercial fields. Sun was the first to successfully target financial houses. Established players in the commercial minicomputer segments, such as DEC and IBM, soon followed. Thousands of workstations can now be found in Wall Street and in commercial banks. Massive purchasing is the norm. Solomon Brother, for instance, bought 450 workstations in 1989 for its traders in New York, London and Tokyo. Many other industries have been lured to the workstation. Hyatt Hotel's information system, for example, now consists mainly of workstations.
3.2 Upstream: Central Processing Unit Technical Segmentation
Because there are tight links between operating system design and central processing unit (CPU) requirements, a key technical distinction is the type of CPU used by a particular computer. Just as applications software often cannot be transferred from one operating system to another, most operating systems are designed to run on a particular CPU line.
Mainframe, mini, and supercomputers often used proprietary architecture, which reduced the potential for multiplatform operating systems standarisation. Among technical workstations and microcomputers, two types of microprocessors were vying for acceptance in 1991: complex instruction set computing (CISC) and reduced instruction set computing (RISC) CPUs. CISC CPUs, such as Intel's 80386 and Motorola's 6040, tended to be cheaper but slower. The Intel and Motorola designs had become the product standards for CISC computing, and the two companies were the major source of CISC microprocessors. RISC CPUs were challenging CISC technology in 1991. The most popular RISC CPUs were produced by Motorola, MIPS, and Sun, while HP, IBM, DEC, and several other firms also offered proprietary designs. A RISC standard had not yet been established.
RISC technology is based on the concept that "simpler is better". Computers spend most of their time operating a limited number of available instructions, so that an architecture that efficiently uses only the most common instructions results in better computer performance. A RISC CPU has a smaller instruction set than that of CISC.
RISC processors have substantial advantages over CISC. First, they sometimes support more complex applications. Second, their speed in operating applications may be double that of CISC. Third, new generations of RISC CPUs can be developed and brought to market more quickly than CISC owing to the simpler design.
RISC architecture also has drawbacks. Because most existing software applications for microcomputers were written for CISC technology, many major vendors such as IBM and Compaq were reluctant to make a major switch to RISC. In addition, RISC based computers need two or three times the memory of a CISC based computer to operate the same software, which increases cost. Moreover, writing a program in RISC often takes longer than in CISC, although compiler advances might address this issue.
More expensive technical workstations tended to be built around RISC CPUs, while microcomputers and lower-end technical workstations usually used CISC. Although most experts predict continued demand for CISC processors, RISC posed a major challenge to CISC as its prices declined and application software became available. Vendors planned to produce hybrid RISC/CISC workstations, but many predicted that RISC would become the dominant microprocessor by the end of the decade. Even in 1991, a CISC 80486 based top-of-the-line system was priced at the same level as some RISC-based workstations with higher performance. Although only 330,000 RISC chips were sold in 1991 (total sales of less than $1 billion), compared to more than 20 million Intel chips, sales growth of RISC-based systems was expected to grow rapidly.
RISC manufacturers followed two licensing strategies in 1991. Some firms licensed their processor architecture to as many other firms as possible, hoping to become the industry standard. Players of this type include Sun, MIPS, Motorola, Intel, and Hewlett Packard. Other firms produced proprietary architecture that they did not license to other firms. IBM, Pyramid, and Intergraph were the major players in the second group. IBM, though, began to change its historic proprietary strategy in 1992, licensing rights to use the chip to Bull and Wang. (IBM also purchased a 5%-10% stake in Bull.)
Table 5 following lists RISC market shares in 1990, while the appendices at the end of the report list report semiconductor sales and market shares in 1990.
Table 5. RISC market shares in 1990
Source: Economist (1990)
3.3 Social Cross-Stream
Public agencies comprised a major market for Unix-based systems in the United States, although their relative importance was declining as the commercial segment grew. For instance, the U.S. Air Force and many major American cities purchase workstations. AT&T was a supplier for military use.
In Europe, the Commission of European Communities (CEC) supported Unix by requiring "open systems philosophies" in the public procurement contracts of their member countries. The CEC itself purchased about 1,000 workstation per year, buying from most European companies.
In Japan, the Sigma project aimed to enhance the position of Japanese companies in the development of software. Participants in the project, which include Japanese firms and subsidiaries of US multinationals, provided most of the $192 financial support, while public agencies provided organisational assistance. The project was expected to be a driving force toward standardisation and to affect the market in the Pacific Rim.
Several public and quasi-public bodies address operating system standards in the United States, Europe, and Japan. American bodies include the National Bureau of Standards, POSIX (Portable Operating System Interface), and the Corporation of Open Systems. These bodies lack authority to set a standard, and act instead by publishing technical specifications with which manufacturers may comply voluntarily. A good meeting the specifications will then be compatible with other compliant products. For instance, a user can expect to connect two hardware components that are "POSIX-compliant". Manufacturers' representatives participate in most such agencies.
3.4 Product Cross-Stream
Complementary Products: Application Software
A computer can gain broad market access only if software is available for it. So popular are some software packages, such as the Lotus 1-2-3 spreadsheet, that a machine can lose sales only because the programs are not available for it. Although the Unix's application library is much smaller than that of DOS (and the DOS-compatible OS/2), every major vendor of microcomputer software has started adjusting its products to the Unix environment. Microsoft, Ashton Tate, Borland, and Lotus Development Corporation had released or announced Unix versions of their most popular software applications to operate under the Unix operating system. These applications include Lotus 1-2-3, QuatroPro, Excel, Microsoft-Word, Wordperfect, Multiplan, dBase, and Paradox. Ironically, Microsoft is a major supplier of Unix software, despite having codeveloped OS/2 (Operating System 2), a major Unix competitor. Major statistical packages such as SAS were also available in Unix versions.
Local Area Network (LAN) developers had also begun to take advantage of Unix portability. Novel developed Portable Ware to operate under Unix. Hewlett Packard and Microsoft developed the LAN Manager/X, a Unix version of the OS/2 LAN Manager/X, that enabled communication between OS/2 and Unix-based file servers. AT&T released LM/X, a Unix tool for the integration of DOS, OS/2, and Unix into a single network. The integration was not yet complete, however, and some application programs operated at best slowly when linked across operating system.
The multimedia market also promised major growth in the 1990s. DEC, for instance, planned to introduce multimedia hardware and software that would let users of LAN-attached PCs capture, store, retrieve, display, edit, print, and distribute full-motion or still-frame video with accompanying audio. Whether multimedia products would use Unix and RISC, or whether other operating systems and chips (such as Digital Signal Processing microprocessors) was an open question.
Substitute Products: Operating Systems
Proprietary operating systems, such as IBM's S/370 and DEC's VMS, provide the traditional substitutes for Unix in the mainframe and minicomputer product segments. OS/2, DOS, and Microsoft-Windows were the major competing operating systems alternatives to Unix for microcomputers. Early predictions were that OS/2 would dominate, backed up by IBM and Microsoft, but it had not lived up to its promise. Windows, which allows some degree of multitasking and is compatible with DOS applications, now appeared to provide the primary challenge to Unix. Operating system market shares are listed in the following table, while characteristics are summarised in an appendices at the end of the report.
Table 6. Worldwide Operating System Market
1988 actual 1993 forecast
($121 billion) ($185 billion)
DOS, OS/2, Windows 26% 29%
S/370 (IBM) 24% 19%
Unix 9% 19%
VMS (DEC) 5% 6%
Mac 3% 5%
System/3X (IBM) 3%
AS/400 (IBM) 4%
Others 30% 18%
Total 100% 100%
Source: Computer Technology Research Corp. (1990)
The most popular Unix versions in 1991 included System V Release 4 (SVR4), AIX, Ultrix, Xenix, and BSD 4.3. AIX and SVR4, supported by the OSF and UI respectively, were expected to challenge for leadership. Their 1988 market shares as well as predictions for 1992 are given in the table below.
Table 7. Unix Market Shares
1988 actual 1992 predicted
AIX, OSF/1 (IBM, OSF) 1% 41%
SVR3, SVR4 (AT&T, UI) 24% 34%
BSD 4.3 (UC Berkeley) 39% 16%
Xenix (SCO) 12% 3%
Ultrix (DEC) 2% -
Others 22% 6%
Source: Computer Technology Research Corp. (1990)
Unix-related revenue realised by computer sector firms between 1986 and 1990 is listed in the appendices following this report.
All major computer market players, including IBM, DEC, HP, and AT&T, and Unisys offer Unix-based systems.
A Big Computer Company (ABC) was minor player in this category. ABC had developed proprietary RISC technology and its own variant of Unix.
International Business Machines (IBM). IBM was the leader in the overall computer market, with total sales exceeding $69 billion in 1990, but it's profitability was declining. IBM was very strong in the mainframe market, where it had a 70% market share. The company was also strong in the minicomputer segment, and entered the microcomputer market in 1984 with the IBM PC. This entry led to the standardisation of the DOS operating system and Intel's 80x86 microprocessor series. IBM was a founding member of the OSF, and its AIX operating system provides part of the base for the Foundation's Unix development efforts. IBM had also recently initiated a venture with Apple and Motorola to codevelop an operating system for RISC-based systems.
IBM was the most vertically integrated computer vendor, producing its own dynamic random access memory chips (DRAMS). The company traditionally emphasised proprietary products and rarely licensed technology to other firms (although it had planned to license older memory chip technology in the failed US Memories alliance), but licensed rights to its RISC technology in 1992. IBM planned to rely on its own RISC technology. IBM was often slow to introduce new products, although it recently had attempted to speed product commercialisation. The company's main strengths included an unmatched global sales force, high investment in R&D ($6.6 billion in 1990), and significant leverage over software developers.
IBM pioneered RISC architecture during the 1970s, but the RT microcomputer was a failure when it was introduced in 1985. The overpriced and underpowered system captured only 2% of the market the first year, and IBM withdrew it. The company aggressively reentered with the RS/6000 family in 1990 and viewed this as a major growth opportunity. Sales of the system, which was designed to be connected to virtually any system, reached more than $1 billion during 1990. IBM enrolled a number of software developers to write programs for the new line based on IBM's AIX version of Unix and many independent developers followed. However, several workstation products promised for 1991 were slow to appear. IBM is described in more detail in the appendices following this report.
Digital Equipment Corporation (DEC). DEC, based in Massachusetts, was very strong in the minicomputer segment, where it challenged IBM with its popular VAX line. A large proportion of DEC's markets consisted of engineers, who use DEC's machine in design applications. DEC introduced technical workstations when workstations began to cut into its minicomputer sales. After three years of development, DEC introduced the successful DECstation 3100 in 1989, which was soon followed by the DECstation 2000 and Decsystem 5400 lines. DEC's major strengths included financial and technical abilities, an enormous installed base of VAX machines, and detailed knowledge of the technical market.
DEC traditionally developed proprietary operating systems, but followed an open system strategy with workstations and was a founding member of the OSF. Having been ousted out of the microcomputer market because it did not use MS-DOS, DEC embraced Unix and develop its own version, called ULTRIX (based on University of California at Berkeley's version). Applications written in ULTRIX's three versions do not operate on VAXs, although the company hopes to develop interfaces. The list of ULTRIX-compatible applications is relatively short. DEC developed a RISC microprocessor. After initially deemphasising the internal product, and licensing in manufacturing rights for the MIPS RISC design, DEC introduced a system based on its own RISC chip in 1992. The chip, named Alpha, would be available for license and Cray planned to introduce a supercomputer that used the microprocessor.
Hewlett Packard (HP). Hewlett Packard offers minicomputers, microcomputers, and technical workstations. During the mid-1980s, HP redesigned its minicomputer line and, when high end microcomputers and workstations started cutting into minicomputer sales, introduced microcomputers and technical workstations. HP doubled its technical workstation market share in May 1989 by acquiring Apollo computers, which lacked the financial strength to support independent growth. HP and Apollo combined equaled 27% market share at the time of the acquisition, only two points behind Sun. The acquisition also enhanced HP's networking capabilities, where Apollo was considered to be strong. HP's technical workstation market share has not kept pace in the past two years, partly because of difficulties in integrating Apollo, which was much more aggressive than the traditionally conservative HP. Nonetheless, HP's sales of workstations and peripherals exceeded $4 billion in 1990, passing the company's minicomputers revenues. In March 1991, HP introduced a new workstation line that matched the performance of IBM units, while being substantially cheaper. HP developed a proprietary Unix operating system and a RISC microprocessor that it licensed to several Japanese and Korean manufacturers. HP was a founding member of the OSF.
American Telephone and Telegraph (AT&T). AT&T was primarily a software firm in the computer industry, but it had positions in the minicomputer, workstation, and microcomputer hardware markets. The company's microcomputer, using its Unix operating system, had little impact. The company was a significant player in commercial markets for Unix-based systems, and it recent acquisition of NCR strengthened that position. The NCR acquisition also provided more internal hardware manufacturing capability, where most AT&T products had previously been subcontracted or assembled from purchased components. In order to position Unix products in the microcomputer market, AT&T announced a Unix version for the microcomputer environment called "Unix-Lite" in 1990. AT&T was a founding member of UI. The company had not developed its own RISC technology.
Firms offering minicomputers using Unix included Amdahl, Prime, Data General, Control Data, and Gould. These firms focused almost entirely on the commercial market segment, although some also sold technical systems. Some minicomputer firms relied on proprietary CPUs but others, such as Data General, had introduced RISC-based systems. Firms in this category had long been pressed by IBM, DEC, and HP, and now faced strong competition from Sun.
Technical Workstation Specialists
Several players in the Unix-based computer market specialised primarily in technical workstation segments, including Sun, Silicon Graphics, Intergraph, and NeXT.
Superbly Quick Architecture Workstation Company (SQUAWC) was a minor player in this category, selling systems designed for architects. SQUAWC used commercially available RISC and Unix technology for its systems.
Sun Microsystems. Sun was the technical workstation leader. It's strength lay in scientific and engineering segments, but the company was broaching financial and other commercial systems segments. Sun was a major player in the product standardisation process. Sun and five of its licensees founded the SPARC vendor council in mid 1980's to promote SPARC architecture. Sun also allied with AT&T to promote SVR4 as the Unix standard. Sun is described in more detail in the appendices following this report.
Silicon Graphics Incorporated (SGI). SGI specialised in technical workstations designed for graphics applications, such as animations, visual simulation and mechanical design. Silicon Graphics is member both of UI and OSF and uses both Motorola's 68020 CISC and MIPS's Risc-based chips. In March 1992, Silicon Graphics announced that is was acquiring MIPS.
Intergraph. Intergraph was the dominant CAD/CAM vendor. In 1987 it acquired Clipper Division from Fairchild Semiconductor company, a producer of RISC-based CPU that Intergraph fabricated in cooperation with Fujitsu. Intergraph targeted the low end of the technical workstation market.
Microcomputer specialists offering Unix-based systems included Apple, Compaq, Dell, and Altos, with Altos being the market leader. Most such systems relied on commercial CISC CPU technology and provided Xenix as the operating system. Most firms offered the ability to operate both Unix and DOS, although the performance of such hybrids had been disappointing so far.
Microprocessor CPU Specialists
Intel's 80x86 and Motorola's 68000 series dominated the market for CISC technology, with Intel being the CISC leader. The two companies had also developed RISC technology, with Motorola having a lead on Intel.
Motorola. Motorola's 88000 series was a new entrant to the RISC market, and a performance leader. Many companies had announced plans to introduce systems based on the CPU. Unfortunately for the company, programs written for the popular Motorola 68000 series CISC CPU, which was the first processor used in workstations, would not operate on the 88000 series. Sun and DEC bought the 68000 CISC chip from Motorola for some of their models, but they use non-Motorola CPUs for their top-of-the-line RISC models. Motorola had recently allied with Apple and IBM to develop a graphically-oriented operating system for RISC-based systems.
MIPS Microsystems. MIPS was an early leader in RISC technology, introducing its first CPU about 1986. MIPS licensed usage and manufacturing rights to more than 20 firms, including DEC, Siemens, Fujitsu, NEC, and Bull. In March 1992, Silicon Graphics announced that it would acquire MIPS, shortly after DEC announced that it would introduce a computer based on its own RISC technology.
Pyramid Technology. Pyramid offered proprietary RISC technology. It was the first company to introduced a RISC-based system and was a niche player with a substantial CPUs market share for high end workstations ($200,000 to $350,000).
Competitors Outside The United States
The American firms IBM, Sun, HP, and DEC were major players in international markets, while no foreign firm had achieved substantial presence in the American market for Unix-based computer systems. In Europe, there was a strong movement toward Unix and RISC architecture. European demand for Unix systems and software accounted for 11% of the market ($2.4 billion) in 1990, and was increasing by 35% to 40% a year. The Japanese Unix market, growing at a 15% annual rate from a $1.3 billion base in 1990, was dominated by HP, Sun, and Sony. Several strong European and Japanese firms seemed likely to become major global players, including Siemens, Group Bull, and Olivetti in Europe, and Toshiba, Sony, Hitachi, Fujitsu, and NEC in Japan. Fujitsu already owned about 44% of the American firm, Amdahl, and had purchased most of the British firm, ICL, in December 1990.
European Electronic National Industry (EENI) was a player in this category. EENI had licensed RISC from MIPS, and was using a Unix operating system that was compatible with OSF/1 for its systems.
Siemens. Siemens intended to be a major international computer manufacturer, but until 1989 70% of its computer revenues were generated in Germany. Siemens then acquired Nixdorf Computer AG, thereby becoming the second largest computer maker in Europe, where Siemens-Nixdorf held more than 16% Unix market share in 1990. Siemens-Nixdorf was also among the top twenty producers of Unix-related computer products in the United States. Siemens was the second largest producer of semiconductors in Europe (after Philips). Current alliance partners included MIPS and Fujitsu.
Groupe Bull and Olivetti. The French firm Bull had a strong European minicomputer position, while the Italian company Olivetti produced microcomputers. Bull was not a major player in the Unix market, but had undertaken microprocessor and Unix alliances. Bull acquired the Zenith computer operations in 1991, thereby forging a minor presence in the United States computer market. Bull sold a 5% stake to NEC in 1991 and a similar stake to IBM early in 1992. Olivetti had a European Unix market share of about 4%. Both companies do not possess their own RISC-based chips, but rely on MIPS.
Sony. Sony entered the Unix market in 1987 with a workstation called NEWS, offering a better price/performance ratio than that of competing American products, and achieved a 16% Japanese market share in 1989. Sony worked with local software developers to write applications for the new computer and soon had over 200 applications in its library. Sony had recently allied with Motorola to introduce a model based on Motorola's 68030 CISC CPU. Sony participated in the U.S. market, where it targeted VARs and CAD/CAM applications, but lacked software applications and achieved a market share of only 1% in 1989. Sony licensed RISC technology from MIPS.
Toshiba. Toshiba was a major semiconductor producer and a leader in portable microcomputers. The company entered the Unix-based technical workstation market in 1988, where it offered low-price products based on SPARC RISC technology. Toshiba had alliances with Sun, HP, Olivetti, and Siemens.
Hitachi. Hitachi was among the world leaders in semiconductors, and a strong mainframe and minicomputer computer vendor in Japan. The company had strong international alliances with corporations such as Motorola and Olivetti. Hitachi introduced a line of workstations based on HP's RISC processors in 1988, and in 1989 began to participate in RISC design with HP.
4. WHAT NEXT?
The standards questions were clear to Sharp, Thomas, and Morini-Stokes. Should they continue to ally with other firms to promote Unix standards or should they pursue standards independently? If they continued to rely on an alliance, should they stay with their current partners or change? Should they promote a merger of the OSF and UI? Perhaps they should simply acquire another firm, rather than ally with it? And what of the same questions for desktop computer standards strategy?
The answers were not as obvious as the questions, even for fictional managers of fictional companies.
And what if the companies were not fictional?
* How would Sharp answer the questions if the company were Sun, rather than SQUAC? Intergraph, rather than SQUAC?
* How would Thomas answer the questions if the company were IBM, rather than ABC? HP, rather than ABC?
* How would Morini-Stokes answer the question if the company were Siemens, rather than EENI? Olivetti, rather than EENI?
Appendix 1. Unix-Related Revenue
Appendix 2. Semiconductor Revenue And Market Share, 1990
Appendix 3. Operating System Characteristics
Some important characteristics of several operating systems are listed below.
System V Release 4 (SVR4): AT&T, UI
• Multiuser and multitasking capabilities
• Distributed computing abilities (Connect microcomputers with technical workstations and mainframe computers)
• Portable: Any application written to Unix can operate in any Unix platform
• Comprehensive network capabilities
• Many Unix versions are not 100% compatible
• Lack of standards
• Unfriendly to end users, although new graphical user interfaces (GUI) such as Open Look and Motif (which are promoted by UI and the OSF) may help
• Not many software applications
• Some deficiencies in security issues
• Requires significant computer/disk capacity
• Excellent for applications that involve integrity of data, high availability, redundancy, time management
• Many additional features, such as remodeled diagnostics, bad block recovery, and remote file access
• Supports symmetric multiprocessing (all processors perform operations simultaneously)
• Multiuser and multitasking capabilities
• System works only with DEC's VAX computer line
Operating System 2 (OS/2): Microsoft and IBM
• Developed by two major players, IBM and Microsoft
• Operates MS-DOS software without any changes
• Operates with Intel's 80286, 80386, and 80486 and microprocessors
• Operates much more complex applications compared to MS-DOS (provides up to 10 MB of memory)
• Multitasking and multiuser capabilities
• Networking ability
• Dynamic linking (efficient use of memory)
• Shipping delays
• Expensive to switch from MS-DOS system
• Lacks software
• Not as user friendly as Apple's Macintosh operating system
• Millions of users
• Many applications available
• Basic factor in microcomputer revolution
• Not multiuser or multitasking
• Limited network abilities
• Does not support complex applications
• Inefficient user interface, less friendly to end user
• Multitasking capabilities
• Compatible with applications software designed for DOS
• Better user interface than DOS
• Low financial cost to purchase
• No multiuser capabilities
• Non-trivial switching costs due to learning time
• Poorer user-interface than Macintosh
• Little software available that cannot operate in DOS environment, so little incentive to switch
AIX: IBM, OSF
• Supported by IBM, OSF
• Compatible with a number of mainframe systems
• Support almost all other operating systems
• Technical excellence (multiuser, multitasking, data base management, connectivity etc)
• Not fully developed
Xenix: Microsoft, Santa Cruz Operation (SCO)
• Suited to personal computers
• Microsoft promoting Windows as PC standard
• Supported by DEC
• Limited number of applications
• Weak in compatibility with other operating systems
BSD 4.3: Berkeley
Often used as the base of proprietary systems.
Appendix 4. IBM's Unix Strategy
Due to the rapid sales expansion of technical workstations and trimming of sales from high-end microcomputers and low-end minicomputer, IBM decided to make Unix-based computer markets a major focus for the 1990s. AIX was IBM's operating system vehicle to enter the market for Unix-based computers. In 1991, IBM offered versions of AIX for many of its popular hardware platforms such as System/370 mainframe computers, RISC/system 6000 technical workstation, and high-end microcomputers. AIX-based systems might cannibalise some sales of other systems (mainly low end mainframes), but the company wanted to offer full line of products in commercial markets for computer systems.
IBM's Unix strategy had several key features, some which differ markedly from its past practices. First, the company planned to comply with all accepted operating system standards and attempt to improve them, rather than pursue nonstandard product development. AIX, its operating system, would support all existing standards and incorporated most key features of previously released versions of Unix. AIX written in the C programming language and offered required features such as multiuser, multitasking, portability, and applications tools. In addition, the program allowed interface to DOS, taking advantage of the millions of DOS users that can access many other systems.
Second, sales would be pursued mainly in the business market, rather than scientific and engineering markets. IBM wanted to target segments such as financial applications and office equipment where growth was expected to be phenomenal. IBM was well-equipped to serve business segments. IBM's reputation for excellent service and vendor reliability was a critical advantage, and the number of applications installed by the competitors was relatively small. In addition, thousands of corporations around the world have acquired IBM's System/370 and all desired connectivity between Unix and their existing systems.
Appendix 5. Sun Microsystems
Sun Microsystems was founded by several young Stanford graduates in 1982, when the workstation industry was virtually non-existent. The company was the leader in the technical workstation industry, with 1990 total sales reaching $2 billion. The growth of the company has been attributed to its open system philosophy and its technical innovation.
Before Sun, technical workstation customers bought computers and operating systems from Apollo and Hewlett Packard. Operating system were strictly proprietary and application software had to be written for the existing operating systems. Thus, customers had limited ability to expand software and hardware.
Sun introduced an open system philosophy by developing a version of Unix, Sun-OS, and then licencing its technology to other manufacturers. Initially, Sun's main sales targets were software engineers and scientific researchers, for whom the key factors were price, performance, and software availability. Sun built its early systems around Motorola's high-performance 68020 CISC microprocessors, used inexpensive "off the shelf" components, and made its system compatible with as much existing software as possible. Benefitting by its technical strength, its willingness to license at reasonable rates, and the absence of giants such as DEC and IBM, Sun grew rapidly. Sun gained a 25% technical workstation market share and, as its revenues jumped from $9 million in 1983 to $210 million in 1986. Sun continues to invest heavily in R&D, with its 14% of sales far exceeding the 8% industry average. In 1989, although sales were growing, the company experienced its first quarter loss in earnings.
In addition to Unix operating system, Sun began to develop its own microprocessor. The company introduced the SPARC CPU based on RISC architecture in 1985. SPARC was more powerful than CISC technology and was tailored to Unix. By 1987, 40 vendors had licensed SPARC technology to manufacture their microprocessors. By 1988, Sun had stopped manufacturing its own CPUs and started purchasing from outside vendors.
Sun used several alliances to promote microprocessor and Unix designs. First, Sun and five vendors founded the SPARC vendor council to promote SPARC architecture. Sun also allied with AT&T to promote the use of Unix based on SPARC technology. Sun's alliances frightened several competitors, leading to the formation of the Open Software Foundation in opposition.
Sun's CEO, Scott McNealy, referred to the company's corporate culture as "controlled chaos". Sun attempted to foster innovation and quick responses by pushing decision making to the lower levels of management. Sun's relations with its competitors were sometimes as chaotic and demanding, and McNealy was not popular with all members of the computer industry. Some industry analysts claimed that the initials "OSF" stood for "Oppose Sun Forever".
Appendix 6. The Emerging Multimedia Market
The multimedia market promised major growth in the 1990s. DEC, for instance, offers multimedia products for local-area network-attached personal computers that let users capture, store, retrieve, display, edit, print and distribute full-motion or still-frame video with synchronized audio. The products will include hardware that lets a PC capture audio and video data as well as software in order to manage the capture and display of audio, still images and full-motion video. DEC will also deliver server software that allows real-time distribution of multimedia information, including the ability to display digital video on a PC from a file on a server. Whether multimedia products would use Unix and RISC, or whether some other operating system and chips (such as Digital Signal Processing microprocessors) was an open question.
Back to the Menu