Last December, I talked to Tom Jarosh, general manager of the AS/400 Division, about the prospects that the AS/400 will become more involved in the Linux and open source communities. While Jarosh, like the AS/400 czars before him, was mum about IBM’s specific plans regarding either Linux or open source, he did hint that the AS/400 Division was waiting to hear from its customers about the incremental value they would get from running Linux natively on AS/400 iron or on the Integrated Netfinity Server for AS/400 (INS) or on the AS/400 bus-attached Netfinity servers that are widely expected sometime in the spring. The AS/400 Division does not really want to support another operating system that could potentially siphon off workloads from the AS/400 itself. It is bad enough that IBM has to endure, much less promote, customers running Windows NT (and in a few weeks, presumably Windows 2000 Server and Advanced Server) on the INS to run applications and systems programs that, by rights, should be on the AS/400, considering the amount of development work that IBM has put into making OS/400 a better operating system than Windows NT or Windows 2000. But Windows NT and Windows 2000 represent the broadest commercial market for such programs, and that means IBM has to offer NT-AS/400 solutions to AS/400 shops, like it or not. By the same logic, the ever- increasing popularity of Linux means that IBM will have to buckle under the weight of enthusiasm. Remember, it took IBM years to put its seal of approval on NT on the INS cards, and IBM only did so after it was apparent that OS/2 Warp and NetWare were dead in the water.
IBM, like every other software development company in the world, has been studying Linux and the open source community that developed it as well as many of the popular Internet programs in use today: The TCP/IP networking protocol, the Apache Web server, and the Sendmail email server are the three obvious open source success stories alongside Linux. Jarosh only told me to “watch this space” about the AS/400 Division’s future Linux and open source announcements. But the little he did say got me thinking. While I would be one of the first people to argue that Linux integration and support on both the AS/400 processors and the integrated and outboard Netfinity servers will soon be a requirement, the possibilities for an AS/400 open source community go far beyond Linux. If IBM truly creates an AS/400 open source community like the one that developed Linux and the other Internet programs, it could go a long way toward revitalizing the AS/400 and even dramatically changing the landscape of the midrange server market.
You would have to be living on a rock on the outskirts of the solar system not to know of the phenomenal popularity of the Linux operating system. Linux is a pseudo- UNIX operating system created by a dissatisfied Microsoft DOS customer named Linus Torvalds. In 1991, Torvalds wrote a baby operating system based on an open source program called MIMIX, which was an early implementation of UNIX for Intel-based PCs. To make a long story short, he was a UNIX open source advocate, and as such, he gave away the code for Linux in the hope that hackers with big egos and free time would tell him how to fix and expand the code. Nine years and thousands of the top minds in operating system development later, Linux is an absolutely credible alternative to Windows 2000 for many print, file, and Web serving jobs and is fast becoming a viable option on the corporate desktop. Linux will also very soon be credible as an alternative to UNIX, OS/400, and Windows 2000 for supporting applications. Linux has about half as many lines of code as Windows 2000, and it will arguably be more stable, too. Throw in the fact that it costs $50 or $80 for a commercial Linux license compared to thousands or tens of thousands of dollars for a Windows NT/2K suite, and it really doesn’t come as any surprise that Linux is the only operating system other than Windows NT/2K that is growing its installed base by anything more than a few percent a year.
The Linux numbers are staggering. IBM has been selling AS/400s since 1988 and has probably not shipped more than 700,000 to 750,000 OS/400 licenses, including all versions and releases. Big Blue has been selling RS/6000 workstations and servers since 1990 and has shipped only about a million AIX licenses. Linux hit 7.5 million licenses installed in 1998, and by the end of 1999, it probably more than doubled that to about 17 million. (As we go to press, the big market researchers are still trying to case the Linux market, which is made more difficult given the open source nature of Linux. Anyone can go to the Web and download it, so no vendor has a good handle on how many Linux licenses are really out there.) While Bill Gates would never cop to it, Microsoft definitely sees those tens of millions of Linux installs as lost Windows NT/2K installs. And so does IBM, which is one of Microsoft’s biggest resellers. (I once saw an estimate by a Wall Street brokerage that suggested that $3 billion of the IBM Software Group’s $11 billion in annual revenues came from reselling Microsoft operating systems.) So if IBM has appeared to be waffling on its commitment to Linux, it has good, sound economic reasons. Pushing Linux means less revenues and, given the stability of the product, probably less money, too, since customers may not need the sophisticated “Cornhusker” high availability (HA) extensions that IBM has developed for Windows NT/2K or the expensive consulting services to install them. IBM is not alone in this quandary. Compaq, Dell, and Hewlett- Packard have exactly the same problem in different proportions because they are all big NT resellers, too.
While Linux is certainly problematic for Microsoft, it would take many, many years to grind down the Windows monopoly, even if the Linux base started growing faster than it already is. As Figure 1 shows, cumulative Linux shipments will probably never catch up to Windows NT/2K shipments, which numbered close to 30 million cumulative licenses shipped in 1998 and will probably break 50 million licenses in 1999. But these cumulative installed base numbers are not necessarily the best gauge of how the operating systems are faring against each other. The number of annual licenses shipped is perhaps a better indicator given the fact that new workstations and servers that replace old equipment get new operating systems and that customers with workstations or servers that are one or two generations behind also tend to keep their operating systems relatively current (especially if the upgrades are relatively inexpensive). My best estimates for 1999 through 2001, which are shown in Figures 1 and 2, assume that Linux becomes a true alternative to Windows NT/2K because Windows 2K is expensive and buggy while Linux remains less sophisticated but more reliable for print, file, and Web serving. As Figure 2 shows, Linux will almost certainly not reach parity with Windows NT/2K in terms of annual shipments any time soon, even under the most aggressive Linux scenarios. But the gap separating the two operating systems will be smaller than Microsoft would like if my Linux optimism and
Windows pessimism pan out. Of course, a lot of things could happen in the next two years. Much depends on how Red Hat, Caldera Systems, TurboLinux, SuSE, and other commercial Linux vendors fare in getting companies like IBM to push Linux as customers pull for it. These companies are going public as fast as possible to build their war chests not only to compete with one another but also to get their message out there in the corporate world that Linux is not only cheaper than Windows NT/2K but also technically better and morally superior.
The Cathedral, the Bazaar, and the Consequences
These latter two items are important distinctions. The secret to the success of Linux is not so much that it is free—using an operating system and maintaining it is orders of magnitude more expensive than buying it—but that some of the best minds in the world are collectively, playfully, and joyfully trying to create an operating system that is more elegant and robust than a commercially created product from Microsoft, IBM, Novell, or any of the myriad UNIX vendors. The key to unifying UNIX was to get the vendors out of the loop, which never happened. That is why there are over a dozen major Unix variants in the world today. Moreover, Linus Torvalds discovered that the open source community was the best way to complete a very large software development project with a minimum of resources in the shortest amount of time.
It is this playfulness and ego, says Eric Raymond in his seminal essay “The Cathedral and the Bazaar,” that makes open source projects work where big commercial software development projects get bogged down and fail. (You must read this paper if you haven’t already. And if you have, read it again. You can get the latest update of it on Raymond’s home page at www.tuxedo.org/ ~esr/writings/cathedral-bazaar.) Simply put, Raymond says that the way operating systems like OS/400 get built resembles the way cathedrals are built: with a few architects, a handful of masons, and lots of disinterested and bored laborers who nonetheless have families to feed and mortgages to pay. Linux and other open source projects are developed by people who are competing to sell their ideas to their peers as the right way to build a product, working over the Internet to create what Raymond calls a “great babbling bazaar of differing agendas and approaches, out of which a coherent and stable system could seemingly emerge only by a succession of miracles.” And yet, quite obviously, it works. Among other tenets, open source development relies on letting anyone who is interested see source code and suggest changes to it, releasing source code early and often, and treating users as debuggers. These are anathemas to monolithic operating systems like OS/400.
But here are two things to consider. First, open source works when applied correctly. (The fact that the Apache Web server is not yet available for the AS/400 shows that IBM has not really gotten open source religion yet and is trying to control that which, by its very nature, will slip out of its control.) Second, no company can afford to pay for the quality and quantity of programming talent that a group of self-selected, motivated propellerheads working through an open source community will do for free, or next to free, for the joy of it.
Small wonder that Tom Jarosh is thinking hard about how to create an AS/400 open source community. Exactly what that means remains to be seen. For starters, IBM could get Apache and Sendmail working on the AS/400. Although it doesn’t currently plan to offer Linux on AS/400 iron, the RS/6000 and S/390 lines already have native Linux beta implementations. But more importantly, IBM should focus on using open source talent to expand both RPG and Java programming for business applications. There are almost a million RPG programmers in the world and over a million serious Java programmers, and at least a few thousand of these are serious AS/400 hotshots who want to see the AS/400 prosper just because they love the box. And although it seems preposterous that IBM would offer open source licenses to OS/400 and its DB2/400 database, that would be a truly stunning development that would prove once and for all that IBM is dead serious
about expanding the reach of the AS/400 in the midrange. It would also upset all of IBM’s business models, so don’t expect it anytime too soon.
What AS/400 customers may really need is for an intrepid bunch of AS/400 RPG programmers to create an open source RPG compiler that works with the DB2 implementation for Linux. While UNICOMP and CrossWorks have AS/400 application rehosting environments that will almost certainly support Linux, they certainly are not open source programs. An open source OS/400, which is what an RPG-DB2-Linux combo would really be, could be the one thing that wakes IBM’s management up to the fact that the AS/400 must compete more effectively against other midrange platforms. With close to a million of you RPG propellerheads, there are probably thousands of you with the talent to create an open OS/400. So what’s stopping you? If Linus can create Linux, why can’t one of you create OS/FourHundrix? You may not make a lot of money, but then again, Red Hat has a market cap of more than $15 billion. I’ll help if you do.
REFERENCES AND RELATED MATERIALS
Eric Raymond’s home page: www.tuxedo.org/~esr/writings/cathedral-bazaar
150 M
120 M
90 M
60 M
30 M
0 M
Cumulative Win NT/2K and Linux Shipments
Linux
1995 1996 1997 1998 1999e 2000e 2001e
Sources: IDC, Dataquest, Vendor Reports, Midrange Computing Estimates
Win NT/2K
1994
0.7 M 1.8 M 6.0 M 15.0 M 29.6 M 52.2 M 85.8 M 134.3 M
0.1 M 0.5 M 1.5 M 3.5 M 7.5 M 17.0 M 35.0 M 65.0 M
Figure 1: Windows NT/2K and Linux have the most cumulative license shipments in the history of server and workstation operating systems.
Annual Win NT/2K and Linux Shipments
Linux
50 M
40 M
30 M
20 M
10 M
0 M
1995 1996 1997 1998 1999e 2000e 2001e
Sources: IDC, Dataquest, Vendor Reports, Midrange Computing Estimates
Win NT/2K
1994
0.7 M 1.2 M 4.1 M 9.0 M 14.7 M 22.6 M 33.6 M 48.5 M
0.1 M 0.4 M 1.0 M 2.0 M 4.0 M 9.5 M 18.0 M 30.0 M
Figure 2: The number of annual Linux shipments could approach those of Windows NT/2K in the next few years.
LATEST COMMENTS
MC Press Online