What's So Open About Open Systems? (And does it matter?)
There's much confusion about just what "open system" means, and argument about whether or not specific architectures and operating systems are really open. Is UNIX, with its multiple implementations, an open OS? Is Windows NT? John Boyd talks with vendors, market researchers, and users about this fundamental issue.by John Boyd
The terms "open computing," "open environment," and "open systems" have been bandied about the industry for more than a decade. Yet they remain as murky today as they were 10 years ago. To understand the current market situation with regards to open systems, we need to take a look at how we got here.
The concept of "open systems" was popularized in the mid-1980s by Sun Microsystems, a workstation manufacturer founded in 1982 in Mountain View, California. Sun sought to contrast its adoption of off-the-shelf components and industry standards (such as UNIX, Ethernet, and Multibus) with the solidly proprietary systems of its workstation rival Apollo Computer (now part of Hewlett Packard) and the mainframe and minicomputer manufacturers such as IBM, Digital Equipment Corporation (DEC), Fujitsu, and Hitachi.
As a differentiating strategy, "open" was marketed with blazing success by Sun, which went on to aggressively portray the rival proprietary systems as "closed." Users and developers immediately took to the idea of "open systems." Working with industry standards made software development and network management easier, and they were less at the mercy of a single vendor's ups and downs.
Open competition
"Open systems" hit the mainstream when influential users of the stature of General Motors, Boeing, Ford, and various US and European government agencies began insisting on compliance with international standards in their tenders. With a hot Sun burning its competitors, rival computer manufacturers also began adopting some of the trappings of open systems, particularly the UNIX operating system.Other manufacturers also co-opted the term "open," which degenerated into a buzzword. The term "open" soon became fuzzy, even meaningless. UNIX, the multitasking operating system (OS) conjured up in AT&T's Bell Labs a quarter of a century ago, eventually became synonymous in most business users' minds with "open" -- although the various UNIX platforms, even today, are rarely "open" to working smoothly with each other (at least not without much sweat, frustration, and talent). As Hideaki Nishimura, a computer systems section manager with pharmaceutical producer Nippon Roche in Kamakura, puts it, "I don't believe it when computer companies tell us their equipment can realize an open environment. There are differences between Sun, DEC, and HP that (negatively) influence our networking devices and parameters."
Like many other corporations, Nippon Roche decided that wrestling with different systems and support tools, and coaxing them to all work together, was a task it would rather do without. "We decided," says Nishimura, "that we should standardize on HP/UX" (Hewlett Packard's version of UNIX).
That's an understandable move, say analysts, but one that, in effect, closes the system. "With all these vendors, and now Microsoft/Intel, fighting each other for market share, and all saying they are open, cost-effective, and offer-attractive solutions, users are not able to decide what is the best solution," observes Junichi Saeki, a senior analyst at market researcher Dataquest Japan. "So, some experienced users are deciding to standardize on a single system. Going with a single provider like Sun, HP, or DEC with Windows NT removes all the problems. The most comfortable situation becomes the best choice for the customer."
What flavor of UNIX?
The problems that continue to plague UNIX today originated during the '70s and early 1980s, when AT&T liberally licensed out the source code to academic institutions and later to computer manufacturers. While this helped disperse the code far and wide, it also resulted in the creation of many flavors of UNIX, since licensees were free to tinker with the code as they pleased.AT&T, however, did maintain tight control over the registered name "UNIX," which explains why UNIX-based computer manufacturers renamed their own versions. That is why today we have exotic names (often ending in IX or UX) for these "nonstandard" standards, such as AIX from IBM and HP/UX from Hewlett Packard.
After the breakup of AT&T in 1985, the company tried its hand as a computer manufacturer and maneuvered to capitalize on the popularity of UNIX. To bring some order to the fractured market, AT&T teamed up with Sun, with the intention of unifying the industry under one UNIX, to be developed jointly by both companies. This was seen by many competitors, though, as nothing more than a blatant attempt to control the technology and give AT&T and Sun an edge in getting products early to market -- an ironic turn of events for a supposedly "open" OS.
Instead of the industry becoming unified, then, it split down the middle into two warring camps: UNIX International (UI) and UNIX Systems Laboratories (USL), led by AT&T, along with Sun, Fujitsu, Toshiba, NEC, and others on one side, and the Open Software Foundation (OSF) with members like IBM, DEC, HP, Hitachi, and Sony on the other.
Eventually, after a distinct lack of success, AT&T retreated from the computer arena, and the UNIX wars petered out. The UI members eventually ended up joining OSF. AT&T ultimately sold the UNIX name and code to networking giant Novell, which in turn recently passed this hot potato on to the Santa Cruz Operation (SCO), the largest supplier of UNIX systems on the Intel x86 architecture.
Sun, which also acquired the rights to the UNIX code, has gone on to develop its own 32-bit UNIX-based multiprocessor (up to 64 CPUs) enterprise operating system, dubbed Solaris. Sun also is now an active board member of OSF.
Open software
Jim Curtin, managing director of OSF's Pacific Operations in Tokyo, says that OSF is working with most of the world's major hardware manufacturers -- and a number of software vendors, including Microsoft -- to enable the broader interoperability of different systems. This would be accomplished through a middleware infrastructure technology called Distributed Computing Environment (DCE), which is a software technology that runs on mainframes all the way down to the desktop. DCE offers common security and naming/directory services.After some twists and turns, OSF's members, which today also include users in the Fortune 500, agreed that what the market was actually asking for was not just portability (the ability to run on various CPU platforms), but rather interoperability (getting systems to work or operate together). "Each customer base has bought into the different UNIX kernels (the core part of an operating system)," says Curtin, "so that a vendor cannot just switch to a new kernel. This is where DCE comes in: to help them interoperate."
Meanwhile, Microsoft dropped a bombshell on the UNIX world in 1993 when it launched its own multitasking, multiprocessor (up to 8 CPUs), 32-bit operating system: Windows NT. With the full support of chip-maker Intel, Microsoft aims to expand its empire from single-user PCs, to cover the entire corporate enterprise. After a slow start, NT is gradually gaining ground. Dataquest estimates that NT will rival UNIX for market share by 1999.
Since the release of the second upgrade, NT version 3.51, in the US in June 1995 (the Japanese version was slated for release in mid-January 1996), Microsoft claims that shipments to new users have more than tripled. In Japan, NT is also gaining in popularity, says Satoshi Uchiyama, director of business consultants Gartner Group Japan KK in Tokyo. "The biggest reason is because of the mass of PC users: they can more easily use the friendly NT. UNIX is still difficult."
Many UNIX hardware vendors, too, are offering NT as an alternative choice alongside their UNIX systems. They are touting this move as a "more open" approach than if they were to merely offer UNIX alone (as Sun still confines itself to doing).
DEC, which provides the only 64-bit rendition of UNIX, running on its powerful 64-bit Alpha servers, is one vendor that has invested heavily in NT, making it part of its central strategy. HP is following the same track. Shigechika Takeuchi, head of Hewlett Packard Japan's Computer Systems Organization in Tokyo, proclaims that the company has no intention of confining its customers to a single system. Rather, HP Japan is promoting "open systems integration and consultation." "That's why," Takeuchi says, "we are not limiting ourselves to UNIX; as the market accepts NT, we include NT."
Sun, however, remains focused on UNIX, and pulls no punches when it talks about NT. "Microsoft's NT is not open, because it's not available from anybody but Microsoft," says Nihon Sun Microsystems' president, Jay Puri. "Fundamentally, open means choice for customers." Puri points to the large number of companies making up SPARC International, the group of semiconductor manufacturers and others fabricating or implementing Sun's reduced instruction set computing (RISC) designs. (Sun itself does not manufacture chips.) In contrast, in the x86 world, Intel does all it can to squash the efforts of Advanced Micro Devices and other semiconductor manufacturer rivals from competing with it.
Reading the numbers
The argument of numbers, however, can be reversed when it comes to counting the flock of computer companies backing NT. Besides the scores of value-added resellers (VARs) and PC vendors -- including Compaq, Dell, Gateway 2000, NEC, Toshiba, and Fujitsu -- many of Sun's UNIX competitors are now also making NT a major part of their marketing strategy.But it's a mistake to take comfort in such numbers, warns Steve Furney-Howe, a director at Nihon Sun Microsystems. "In a Microsoft world, innovation and product timetables still all depend on Microsoft." Would, he asks rhetorically, NEC have had the recently reported installation problems with Windows 95 if Windows 95 was really an open interface?
An open environment, "where specifications and interfaces are open for all to use," says Furney-Howe, "enables innovation to come from anywhere at any time." To illustrate what unfettered innovation in an open environment can do, he points to the Internet, and to Sun's recent push to open the Net further with its Java programming language.
The Internet model of openness
Certainly it's fair to say that the Internet's dynamism and exponential growth owe much to the widespread innovation that has been made possible by the openness of the Net: the openness of its structure, end-user software, and underpinning philosophy (with the latter accounting for a purposeful lack of central control). This openness contrasts sharply with the strategy of using tightly controlled proprietary technologies to gain a monopoly domination, as Microsoft has done in the PC market.With Java, Sun is seeking to take the open environment of the Net to a new level. Java is an object-oriented, interpreted language for the Internet that enables programmers to liven up static Web pages by writing dynamic, animated interactive programs called "applets." But more than being cute programs, the same applet will run across multiple platforms: on Solaris, Windows NT, Windows 95, and MacOS 7.5 systems that have Sun's HotJava Web browser installed, for example. The Hot Java browser is available for downloading free at http://Java.sun.com/.
Microsoft has recently demonstrated its own "openness" to market currents by mounting a major strategic turn. In order to make itself "Internetcentric," Microsoft has even decided to license Java from Sun. Moreover, Microsoft is quick to point out that no one has a lock on innovation. "We are not stopping to innovate for a moment," says Charles Stevens, vice president of Microsoft's Far East operations, in Tokyo. "We have a huge incentive to innovate, just like each UNIX vendor has. Okay, we are not totally open 100% -- but then, nothing is."
The UNIX lie
Besides, when all is said and done, Stevens argues, what really matters is the kind of product you deliver to the user. "Anyone can define 'open.' What is important is, what does the customer want?"Microsoft's answer is lowest system costs, software training and support, a large number of applications, portability, scalability (the ability to run on any size of computer), and multivendor interoperability through OSF's DCE. "The benefit from all these is flexibility at a reasonable cost," says Stevens. "And if you agree with this premise, then NT does well in all these categories."
Stevens also takes pains to point out the "lie" of UNIX. "There isn't one UNIX; that's a myth. There are many varieties. UNIX is not binary compatible across any platform, so software developers have to recompile and test products for different systems -- and probably sell them in different boxes." In the case of NT on the other hand, Stevens notes, "you need produce only a single application in one box."
Bigger problems
And so the arguments rage. Yet for some users, the question of "open systems," and whether NT is less or more open than UNIX, pales to insignificance when they are faced with more fundamental problems.Ian Hayden, manager of Information Systems at Agfa Gevaert Japan Ltd., in Tokyo, is one such user. He heads a six-man team managing a Novell NetWare 3.1 network of scores of Compaq and Macintosh PCs, an NT server, and a Sun workstation -- all coming off a local IBM AS/400 mid-range computer that provides the link to the company's host computer in another part of Tokyo. Open systems "is not so much the case of hanging all these different systems on a wire and getting them to run nice and friendly, although that's what some people think," says Hayden. "In a sense, we've got that. The key feature of an open system, for me, is the operations management side of the system."
In other words, Hayden is more bothered with finding the human resources to manage the different systems, and still keep within his budget. "I'm not looking at this from a UNIX versus NT point of view; I'm looking at it as AS/400 versus the rest. My open systems problems are of quite a different nature: resources."
Gartner Group Japan's Uchiyama may have put his finger on the central issue when he says the word "open" is not so important anymore. What is important is whether the systems in question offer the important components that compose the concept of "open": portability, interoperability, industry standards, and choice. "When you cannot clearly define 'open systems'," Uchiyama points out, "then a system's features, its reliability, its cost, and the choices it provides the particular user -- these become more important criteria for making a purchase."
