Understanding how these divergent approaches to creating operating systems developed requires a look at how the worlds of PCs, workstations, and servers came to collide [see "The changing workstation"]. PCs, at the lower end of the computing spectrum, have been dominated by Microsoft's operating systems since the 1980s. Until 1993, however, workstations were the domain of Unix operating systems. At that time, PCs and workstations were very different.

PCs were in the main used for office applications (as they mostly are today), whereas workstations were designed for technical and scientific tasks. In terms of performance, workstations stood between the PC and the minicomputer, having more RAM and better floating-point computation and graphics capabilities than PCs but a less powerful processor and less in the way of memory than a minicomputer.

Like PCs, workstations can serve a single user. Yet like minicomputers, which typically served several videodisplay terminals in a department, workstations can also act as servers in network environments. In the early '90s, PCs based their single-microprocessor architecture on complex-instruction-set computing (CISC) devices belonging to the 80X86 family from Intel Corp., Santa Clara, Calif. Workstations, on the other hand, used reduced-instruction-set computing (RISC) processors made either by the workstation manufacturer or by independent processor designers and manufacturers, like MIPS Technologies Inc., Mountain View, Calif., and Motorola Inc., based in Schaumburg, Ill.

Just as PCs and workstations at first focused on different kinds of end uses--applications, that is--so, too, did their operating systems. The PC's DOS and Windows operating systems were designed to run on Intel chips alone, and Windows had a graphical user-interface added to it to woo the computer novice. Unix, developed in the early '70s at the then American Telephone & Telegraph Corp., was not intended for beginners: it was built by and for computer professionals who were comfortable with programming and software architectures.

Furthermore, to move Unix from one type of microprocessor to another, users could recompile its source code, which was written in the C programming language and could be licensed from AT&T for a reasonable fee. (AT&T had been restrained by U.S. Federal courts from entering the computer market itself.) It was not long before Unix found its way into the university community, running on a variety of platforms.

The source code for Microsoft's operating systems, on the other hand, remains proprietary and closely guarded to this day. A few researchers and business partners have had access to it, but only under extremely strict nondisclosure agreements. Microsoft alone may alter it or move it onto a hardware architecture other than Intel's.

Thus it was that in the late '70s, Bill Joy and other students at the University of California, in Berkeley, were free to create an improved version of Unix--the Berkeley Software Distribution, or BSD.

The start of a new era

Joy's work on Unix BSD drew the attention of Vinod Khoda and Andy Bechtolsheim, graduate students at California's Stanford University who in 1982 were transforming a university project on networked computers into a start-up. That was how the Stanford University Networked (SUN) workstation gave rise to Sun Microsystems Inc., in Mountain View, Calif. Khoda and Bechtolsheim hired Joy to take charge of the workstation's software and, in so doing, thrust Unix into the commercial arena. Sun's successful use of openly available software technologies and standards, such as Unix and the IEEE 802 group of Ethernet standards, forced earlier entrants into the workstation market like Hewlett-Packard Co., Palo Alto, Calif., and the then Apollo Computer Inc. to abandon their proprietary system software by the late '80s and adopt Unix.

Sun Microsystems would also lead workstation hardware away from the CISC architecture. CISC was at first universally employed in the computer industry, even by Sun in the form of the Motorola 68000 chip family. Then when the RISC architecture was invented in the mid-'70s by John Cocke, his employer, IBM Corp., held off on its commercialization. Nonetheless, academic researchers, led by David Patterson at the University of California at Berkeley and John Hennessy at Stanford, had gotten wind of IBM's work and had started their own reduced-instruction-set architecture projects; UC Berkeley's project was simply called RISC, while the one at Stanford was named MIPS (for millions of instructions per second). In 1987, Sun applied Patterson's work to a workstation based on what it dubbed the scalable processor architecture (Sparc) microprocessor.

RISC processors of the late '80s and early '90s worked faster than traditional CISC chips from Intel and Motorola. Their clock speeds were no greater, but their architectural features, such as the pipelining of instructions, let them work on more than one instruction at a time. What made this possible was a reduced set of instructions, each of which, for the most part, took about the same length of time to execute. Swayed by the superior performance of RISC architectures, in 1988 Microsoft began developing its first operating system capable of working on RISC as well as CISC chips, which would become Windows NT.

A bridge to RISC

Microsoft's new operating system could in theory be used with any processor chip, RISC or CISC, thanks to its use of a special layer of intermediary, or meta-, software between the main operating system and the microprocessor. Dubbed the hardware abstraction layer, or HAL, this software translated NT's commands into the operating instructions specific to each type of microprocessor, and vice versa. With Microsoft's cooperation, a chip vendor could create a HAL to run on its silicon, and Microsoft would help support that implementation.

With the delivery of NT in mid-1993, the software giant now had a 32-bit operating system that would run on the workstations' RISC processors. The very first chip to boot up Windows NT was the new Alpha RISC processor, from Digital Equipment Corp. (DEC), in Maynard, Mass., which also could run Unix and the Virtual Memory System OS, better known as VMS.

Another early NT supporter was the MIPS processor employed by Silicon Graphics Inc., in Mountain View, Calif. (Interestingly, the first release of the operating system was dubbed Windows NT 3.1, which Microsoft explained as a means of coordinating NT numbering with its current PC operating system, Windows 3.1.)

Rather than lose out to the new processor architecture, Intel utilized those features of RISC design that would speed up chip performance without invalidating the on-chip operating instructions used in earlier chip designs or the software made for them. The result was the 32-bit Pentium family of chips, first introduced at about the same time as Windows NT. Thus the so-called Wintel architecture (a label for the PC design formed by combining Windows and Intel), was set to follow up on its success in the PC market by taking on the Unix-dominated world of workstations.

Certainly, the field looked ripe for harvest. Sun's success with Unix had, by the mid-'80s, impressed other computer vendors--notably DEC, Hewlett-Packard, IBM, and Silicon Graphics. They decided to adopt Unix, and set about trying to enhance it further, each of them developing a slightly different variation.

An unfortunate side effect was the fragmentation of the market for Unix applications because the independent software vendors had to customize their applications for every variety of Unix they supported. Instead of having one application that would run on many PCs, the vendors had to have many versions each serving relatively few workstations. By designing for Windows, vendors were ensured of having a wide market for one version of their software, and their development costs were spread over many more unit sales. Then, too, users had to support multiple versions and revisions of Unix, pumping up their cost of doing business. Windows reduced considerably the support cost, and gave them many vendors from whom they could purchase a wide variety of applications.

Even the Unix suppliers recognized that they were dividing up the application market and attempted to unify their Unixes. In 1988, they formed the Open Software Foundation to try to agree upon a common standard for Unix, but it was seven years before they could agree upon their first version of such a standard: the Single Unix Specification. Those who met the specification were allowed to use the Unix 95 brand.

In addition, they supported the Portable Operating System Interface for Unix, or Posix--a set of IEEE standards that define an interface between programs and operating systems--and encouraged the independent software vendors to do so, too. The idea was to reassure the developer of programs that conformed to Posix that they could be ported to Posix-compliant operating systems without much fuss. Even so, additional commands show up in some Posix-compliant Unixes, and when software vendors succumb to the temptation to use them, porting the resulting programs can become a much bigger job than is the case with Posix-only applications.

The adoption of Posix has an unusual benefit for applications. Although conceived of as a Unix interface, Windows NT had Posix compatibility as one of its goals as well. In theory, fully Posix-compliant applications would run either on Unix or on NT; but in NT they would not run as quickly as those written to the NT application programming interface called Win32.

Getting NT's Posix subsystem to work properly turned out to be hard sledding. In 1995, Microsoft turned for help to Softway Systems Inc. and invested in the San Francisco company. But only in October 1998, less than a year ago, was it certified, under the name of Interix, as being a fully compliant Posix/Unix environment for Windows NT [Fig. 1].

Naturally, the vendors of workstation software envied their PC counterparts and found Windows NT's level and unfragmented playing field very alluring. As for the vendors of Unix systems, the temptation to tinker with Unix and thereby gain an edge in the market over their competitors, rather than compromise and adopt a common standard, proved irresistible.

Rumors of its death

To many observers at the time of NT's introduction in 1993, the new Microsoft operating system looked as if it would be the death of Unix. It might have been, too, barring another development, which Microsoft had overlooked: the Internet. While Windows NT was being designed, the company had not yet grasped the Internet's significance and how it would affect system requirements. But by the spring of 1994, finally, "Microsoft was betting that the Internet would be important someday, and we were building support for it into our products," Bill Gates wrote in the preface to the second edition of his book, The Road Ahead [see To Probe Further].

While Gates's company had previously acknowledged some value in interconnecting PCs in some way, it viewed the computer users' world as an archipelago of independent states. Each desktop was an island fiefdom, with lines of communications to other similar regimes. Granted, this local-area-networked world also contained warehouses--databases and file servers for storing records and documents--but they functioned as peripherals to the PC. The user's desktop was the center of each state, and each center needed to be fully loaded with applications it could run on its own.

The PC revolution had, after all, been sparked by the desire for freedom from terminals that shackled users to a giant mainframe and from the whims of the Information Systems department. Consequently, Windows NT started out as a more powerful operating system for the desktop, a VMS for a personal VAX minicomputer, as it were.

Unix, as the operating system for workstations, had also been targeted at a single user who needed to run a particular type of very demanding application, such as an electrical engineer who needed to run a hardware simulation. But the world view that had nourished it was not as single-minded. Having begun its life at AT&T on a shared VAX minicomputer from Digital Equipment, Unix accommodated many users simultaneously on a single system. That meant running several applications and juggling many different inputs and outputs, all at once.

Then Unix found its way into the universities, which it could serve in two essential ways. Thanks to the availability of its source code, it could be used to teach how operating systems (and programmers) work. Access to the source code also simplified deploying and maintaining the operating systems in working systems. Since many universities participated in the Internet's progenitor, the Arpanet, it was natural for those institutions to employ Unix in the servers they attached to that net. (Harvard, the Massachusetts Institute of Technology, Stanford, and the University of California at Santa Barbara had been participants since 1970.)

Thus by the mid-'80s Unix was driving the Internet, and Sun's university-born systems were already deployed as servers all over the Net. No wonder, then, that Sun took the view that "The Network," and not its component servers, "is the Computer."

Marketing muscle

Unix had captured the hearts and minds of technologists and academics, but it faced stiff competition in business circles from Microsoft, which, in terms of its successful penetration of corporate computing, had become the IBM of the '90s. Purveyors of the various Unixes were battling with each other as well as with Microsoft, and offering myriad configuration options--a boon to technologists but confusing to executives who just wanted black-or-white choices.

With NT, Microsoft could in 1993 position itself as a clear-cut choice and a more modern operating system. It offered the software with essentially only two options: NT for workstations or NT for servers. Further, because one of NT's chief designers, David Cutler, had overseen the creation of the VMS operating system for Digital Equipment's VAX line of minicomputers, the NT had a good technological pedigree.

Still, NT was off to a slow start. According to MIT researchers Michael A. Cusumano and Richard W. Selby [see To Probe Further ], Microsoft had predicted sales of a million copies in the first year of the operating system's existence, but had moved only 300 000 by mid-1994. The main reason was that pre-existing 16-bit applications ran more slowly on NT than on Windows 3.1, which also made do with less system memory.

Accordingly, the first major revision, NT 3.5, released in September 1994, perked up the operating system's performance for those legacy applications and demanded less system memory. The upgrade boosted sales to 1 million copies by December.

But what really got the ball rolling for NT was the incorporation of a graphical user-interface. Borrowed from Windows 3.1's successor, Windows 95, it made NT handier to use. The enhancement appeared in NT Workstation Version 4.0, released in July 1996. By November 1997, according to International Data Corp., Framingham, Mass., over 11 million copies had been licensed and in a year that number jumped to 25 million.

Exaggerating somewhat NT's success in traditional workstation markets was its invasion of office systems formerly served by Windows 95. Microsoft positioned NT as a corporate operating system more rugged than Windows 95 or 98, whose underpinnings are, after all, the aged 16-bit MS-DOS. The shift to the NT architecture was to be completed when Microsoft shifted to a single desktop operating system--Windows 2000, the new name for Windows NT 5.0--later this year, but Microsoft has recently hedged that commitment. Further, for the most part, NT is running on Intel Pentium chips, not RISC processors. The only RISC processors Microsoft still supports are the DEC Alpha family, whose architecture is now owned by Compaq Computer Corp., the Houston company that took over DEC last year.

The situation is somewhat ironic. Although Compaq owns the DEC architecture, Intel manufactures these RISC chips which with its own CISC-architecture processors once fiercely competed. Intel's involvement with Alpha means that, in a sense, Microsoft now supports NT's use only on Intel microprocessors, perpetuating the Wintel duopoly.

Underlying strength

From the market figures alone, it is hard to cull those NT systems sold for technological use from those intended for business application. Performance is a better indicator of which systems would best suit technical work. Judging by the performance metrics gathered by the Open Systems Group of Standard Performance Evaluation Corp. (SPEC), Manassas, Va., the systems with the highest throughput and speed for floating-point operations (which typify engineering work) all use RISC processors running Unix software.

That is to be inferred from last year's SPEC95 benchmark figures; while the best test of performance for any individual company is to run the actual applications that their users will rely on, SPEC benchmarks provide some means of comparing performance. The finding holds firm even if the number of processors in a system is only a single chip. The result is nothing to be surprised at, considering that many Unixes are available in 64-bit versions, as are RISC processors, whereas Windows NT is merely a 32-bit system.

The SPEC results also underscore a trend in system design that began several years ago: the use of multiple processors in a single system. The highest commercial performance is achieved by IBM's RISC/6000 SP system, which runs IBM's AIX 4.2.1 version of Unix on fully one hundred ninety-two 160-MHz Power2 processors. (This IBM central processing unit is an architectural ancestor of the PowerPC microprocessor created by Apple, IBM, and Motorola.) Admittedly, this is the extreme in parallel processing, and two-, four-, and eight-processor systems abound that are suitable for simpler server and workstation applications.

At any rate, that is true for the RISC /Unix combination. NT is ill able to support several processors, so multiple Pentium/NT systems are uncommon. One instance is from Dell Computer Corp., in Austin, Texas, which configured its Precision WorkStation 610 with a pair of 450-MHz Pentium II Xeon processors running Windows NT Workstation 4.0--yet a Sun Enterprise 450 with a single 400-MHz UltraSparc II running the Solaris 7 Unix operating system gives the same throughput.

Performance is not the only system yardstick. Features like scalability as well as reliability, accessibility, and serviceability (RAS), can sometimes offset a slight lack of processing muscle. Those virtues, though, are less evident in NT than in commercial Unix, according to research and consulting firm D. H. Brown Associates Inc. With regard to the operating-system scorecard it maintains as part of its system software research efforts and publishes on its Web site [Fig. 2], the Port Chester, N.Y., company concluded that "Windows NT Server 4.0 Enterprise Edition still trails UNIX in every area except for PC client support. Microsoft has made progress in driving visionary features into the base operating system--including built-in transaction processing and strong links between database and Web servers--but NT falls short of matching UNIX competitors for advanced Internet protocols and extensions. NT also continues to lag in scalability, RAS, and system management...."

Support for this view comes from John Kirch, a networking consultant and Microsoft Certified Professional who helped found the UNIX Versus NT Organization. In a Web paper comparing server versions of both systems, he not only provides more information supporting The Scorecard's claims, but also analyzes the costs associated with NT and Unixes. His conclusion: Unix is in point of fact less expensive to own than NT [www.unix-vs-nt.org/kirch/].

In reaction to this onslaught, Microsoft recently admitted its shortcomings in scalability and reliability, but said it plans to overcome them with the Windows 2000 family of workstation and server operating systems. The software has been in beta testing since September 1997 and is due for release later this year. Members of the new family will include support for up to 16 processors and will offer "the highest level of security, state-of-the-art features for mobile users, industrial-strength reliability and better performance...while lowering the total cost of ownership through improved manageability," to quote Microsoft.

The worst kind of competitor

If the software giant had merely to compete against traditional software sellers, its history would portend success for Gates and company. But new opposition has arisen, in the form of a growing movement to freely develop and give away Unix. Boston's Free Software Foundation, which has set itself up as a tax-exempt charity, raises money to support the GNU project, which was started in 1984 by foundation president Richard Stallman.

GNU is a recursive acronym that stands for GNU's Not Unix, a name that recalls the days when AT&T vigorously protected its copyright of the name Unix. Thus GNU is described as a "Unix-like" operating system. And, being Unix-like, GNU is not monolithic; rather, it is composed of a variety of modules, such as a graphical user-interface, an editor, a file management system, and other utilities, which are added to the operating system kernel. The user decides which modules are required and adds them as necessary.

Another significant contribution of the software foundation is the GNU General Public License, a legal document that allows source code to be freely distributed in a controlled way, referred to as "copyleft." A copyleft license states that anyone who redistributes the software, with or without changes, must pass along the freedom to further copy and change it.

GNU has not yet created its own operating system kernel, but a free one exists that has recently set the wires humming. Like the BSD version of Unix created by Bill Joy, the kernel has its origins in academia. In 1991, a student and computer hobbyist at the University of Helsinki, Finland, built a small but complete operating-system kernel that would outperform Minix. Minix is
a Unix clone written from scratch (and therefore free of AT&T royalties) that is still used for software instruction at many universities in Europe.

By combining his first name with the name of Unix, the student--Linus Torvalds--christened the software Linux (pronounced LINNucks). Three years later he had readied Version 1.0 for release. While disinclined to turn the outcome of his hobby into a business, Torvalds did want to get credit for his work and to encourage others to help enhance it. So rather than lock away Linux's source code, he distributed it under the GNU General Public License.

The free and open availability of Linux proved seductive to Webmasters and network administrators, who began to deploy it--often surreptitiously. A software module called Samba made Linux look like an NT platform from the server user's perspective. What helped it go undetected was its high reliability; since it rarely went down, it was hardly ever a problem that needed to be investigated.

There were several reasons for the initial secrecy over Linux's use in corporate environs. For one, it was not formally supported by a commercial entity. Informal, on-line groups could help with problems and supply fixes and additional modules, but there was no vendor who could be taken to task if things did not go right, or who could be put under contract to supply support for the system.

A second reason is the fact that those promoting the software were a self-described community of hackers, iconoclastically looking to change the way software was created and sold. A characteristic of that community is its tendency to combine business with other issues. At the Free Software Foundation (FSF) Web site [http://www.fsf.org/], for instance, information is available about not only the GNU project but also, on the site's personnel pages, the tongue-in-cheek "Church of Emacs," complete with a hymn and a resident saint--"Saint IGNUcius." (Emacs is a text editor for Unix created by FSF founder Stallman, masquerading here as the saint.)

In a more serious--although still iconoclatic--vein, the site points out the plight of Taslima Nasrin, exiled from Bangladesh after being physically threatened and officially charged with the crime of blasphemy because she criticized the persecution of Hindu citizens and the injustice of Islamic law. It also briefly presents Stallman's view on persecution by religious organizations.

Two factors have weakened corporate objections to Linux. One is the performance achieved in various academic and military projects. Consider the Beowulf project, a NASA initiative sponsored two years ago by the High Performance Computing Systems Group, part of the Jet Propulsion Laboratory (JPL), in Pasadena, Calif. Beowulf used Linux to tie together a cluster of 16 PCs--a so-called Pile of PCs--for less than US $50 000; on a complex gravitational simulation problem, the system was able to operate at a sustained rate of 1.25 gigaflops. Other undertakings include the FlowNet high-speed network project of JPL and Cedar Technologies Inc., Edina, Minn., and a prototype Linux embedded system being built by The Mitre Corp., Bedford, Mass.

The second factor is the adoption of Linux by vendors pursuing a new business model. Caldera Systems, Pacific HiTech, Red Hat Software, Suse Holding, and VA Research are all distributors of Linux that prefer not to charge for the software itself, but instead collect fees for its documentation and support.

Coming of age

Back in September, Intel, the hardware side of the Wintel duopoly, and Microsoft arch-rival Netscape Communications Corp., Mountain View, Calif., both announced that they would make equity investments in Red Hat Software, located in North Carolina's Research Triangle Park. The idea was to enable Red Hat to establish an enterprise computing division to beef up Linux support. This commercial backing has in turn received the endorsement of IBM, which has announced support for Linux not only on PCs but also on its RS/6000 line of workstations and servers.

Compaq Computer, Dell Computer, Hewlett-Packard, and Silicon Graphics are others that will bundle and support Linux with their systems. For all these computer companies, Linux is becoming a viable alternative to Microsoft operating systems as well as to its own versions of Unix. And just last March, Intel and Netscape were joined by Compaq, IBM, Novell, and Oracle as equity investors in Red Hat.

As yet, the copies of Linux in use total about 7.5 million, by Torvalds' reckoning. In marked contrast, shipments of Microsoft desktop operating systems came last year alone to about 80 million, versus half a million copies of Linux, according to a market report released last January by International Data Corp. On the other hand, in a separate report on server operating systems, International Data found that Linux was growing "at a dramatic rate of 212.5 percent, accounting for more than 17 percent of all [server operating systems] shipped." With 1.56 million copies delivered in 1998, NT took second place in growth rate, at 27.2 percent, compared to all other Unixes, which grew only at 4.1 percent. (In all its flavors, Unix is still the server market revenue leader.)

Do open source software and Linux worry Microsoft? Yes, according to two confidential memos written by Microsoft staff engineer Vinod Vallopillil on 11 August 1998. The two are referred to as the Halloween memos (www.opensource.org/halloween1.html and /halloween2.html) because the first appeared on 31 October 1998 and the second on the following day at the Open Software Initiative (OSI) Web site. They were published by Eric Raymond, an OSI founder and software consultant in Malvern, Pa. The source of the leak is not known.

In his memos, Vallopillil wrote: "Linux represents a best-of-breed UNIX, that is trusted in mission critical applications, and--due to its open source code--has a long term credibility which exceeds many other competitive OS's." Furthermore, in responding on 5 November to the leaked messages, Microsoft's enterprise marketing group manager Ed Muth confirmed that: "Linux is a competitor on the client and the server. My analysis is that Linux is a material competitor in the lower-performance end of the general purpose server industry and the small to medium-sized [Internet service provider] industry," adding that Linux also competes with other versions of Unix. The fact that their employer was, at that time, deep into its Federal monopoly trial may have made the company more eager to show it had competition, but the International Data numbers confirm that Microsoft ought to have been concerned.

Should other purveyors of Unix give up on their own versions and bow down to Linux? Not quite yet, according to Linux: How Good Is It?, a report released last month by D. H. Brown Associates. In comparing Linux releases from Red Hat and Caldera Inc., Provo, Utah, against the server versions of Unix from Compaq, Hewlett-Packard, IBM, Silicon Graphics, and Sun, as well as Microsoft's NT server, the report judged that "for meeting the functional requirements of enterprise computing,...the leading conventional UNIX operating systems and Windows NT hold an overall advantage."

While finding that Linux had strengths in some applications, such as entry-level file- and print-sharing systems and Web servers, and as compute nodes in technical computing clusters like the Beowulf systems, the report questions the ability of the open-source community to match the innovation of conventionally developed software.

"Conventionally developed operating systems endured a lengthy and painful maturation process before they achieved enterprise-grade levels of scalability and reliability, while delivering enterprise functionality," says the report, noting, "This is a process that Linux has just begun."

But commercial operating system suppliers are beginning to experiment with the open path. On 16 March, Apple Computer Inc. did so when it introduced its latest operating system, OS X. For the first time, a prominent computer company has made source code for its operating system--not all, but a good portion of it--open to public scrutiny. As a large commercial software developer, the Cupertino, Calif., company is following in the footsteps of Netscape Communications, which opened up the source code for its Communicator Web browser/calendar/e-mail client in February 1998.

As this article was going to press, Brian Valentine, the Microsoft vice president in charge of Windows 2000, said the company was seriously considering publishing the source code for the operating system's kernel. Should that happen, open source code will have gained the momentum it needs to succeed as the preferred method for software development.

To probe further

A History of Modern Computing by Paul Ceruzzi, curator of the National Air and Space Museum's department of space history (MIT Press, Cambridge, Mass., 1998) deals concisely with the electronic computer from W. J. Eckert and J. W. Mauchly to Gates and Joy. This highly readable account also covers the technologies and documents its information thoroughly; the last 70 of its 382 pages are given over to notes and a bibliography.

Two books help penetrate Microsoft's thinking: Microsoft's Secrets, by Michael A. Cusumano and Richard W. Selby (Simon & Schuster's Touchstone imprint, New York City, 1995; paperback, 1998), and The Road Ahead, by Bill Gates with Nathan Myhrvold and Peter Rinearson (Penguin Books, New York City, 1995; second edition 1996). Secrets shows how the company sets about forming strategy and executing it. The Road Ahead portrays the perspectives of Microsoft's leaders. The difference between its first and second editions shows how quickly a perspective can change in the face of technological developments.

The January/February issue of IEEE Software focuses on Linux and includes an "Etherside Chat" with Eric Raymond, as well as articles about the business aspects of Linux and open software and how it is being employed today.

The Operating System Scorecard is published regularly by D. H. Brown Associates Inc., Port Chester, N.Y., as part of its research reports and at its Web site [http://www.dhbrown.com/]. The company also publishes a newsletter, Tech Trends Monthly, covering software advances and their effects on corporate decision making.