Open Source OS/400: A Crazy Idea for Crazy Times

by Timothy Prickett Morgan

IBM's chairman and former CEO, Louis Gerstner, has been fond of saying over the years that operating systems are no longer control points in the IT industry. Hence, IBM's very loose grasp on the intellectual property behind OS/400, AIX, and MVS (sarcasm intended). IBM has historically retained tight a control over its operating system code, as if the open source revolution has not swept over the IT landscape. Maybe it is time to change that. Maybe it is time for open source OS/400.

This is not the first time I have had this thought--many of us have--and I am certainly not the first person to write about an open source version of OS/400, which has been referred to as Open Source/400, as a possibility. Industry pundits and analysts, not to mention people within IT suppliers with closed operating systems like IBM's, have all probably been mulling this over since the late 1990s, when it became clear that open source technologies were breaking the grip that vendors would have with proprietary platforms like OS/400.

Gerstner was, of course, correct when he said that operating systems are no longer the control points--a cynic might say choke hold--that they once were. This has less to do with Linux and open source than it has to do with the prevalence of Unix operating systems in the late 1980s and throughout the 1990s as a cheaper, more-open alternative to proprietary platforms like OS/400, MVS, VMS, and myriad other platforms that the kids in college have never heard of. Like many revolutions, it takes a few turns to actually foment change. The first volley in the open source revolution was really about open systems. Open systems meant Unix, and, specifically, it meant that the Unix vendors would agree on a set of APIs and then implement them, in closed-source fashion, on their respective platforms. Eventually, when Unix took off and explosively came to account for half of total worldwide server revenues, vendors of proprietary operating systems had to reconcile themselves to the fact that if they did not support the Unix APIs, no new applications would come to the platform and existing applications might not be modernized.

Hence, in the mid-1990s, everyone was talking about the SPEC 1170 APIs, which are the core Unix APIs that all modern operating systems support. OS/400 supports all but a few of the SPEC 1770 APIs, ones dealing with supercomputer applications, and MVS--now called z/OS, after being called OS/390 for a few years--actually has all of the SPEC 1170 APIs and the POSIX operating system interface specs that would technically allow it to be branded as Unix. Believe it or not.

But being an open system is not the same as being open-source. Open source is the most ridiculously open of opens. Everyone can peer inside the code. Anyone can take that code and use it as he sees fit. Anyone can, under various open source licensing models, modify that code to make it better, so long as those modifications are made available to others. Open source is a development method and ethic that is based on the idea that the best code, usually written by the best programmers in the world, and on their own nickel, for kicks and ego trips, will eventually get into the program, and that this will result in a better program. It is simultaneously a democratic and an elitist approach.

Open source is as much an anathema to the software development methods of IBM, Oracle, Microsoft, and the other software powerhouses as the AS/400 and its predecessors have been something of an enigma among the several families of servers that IBM has sold in the past three decades. But this, alone, is not a good enough reason to take OS/400 open-source. The best reason to do it is that it makes good business sense. Operating systems may not be a control point because of the adoption of Unix standards, but hardware and operating-system-platform distinctions are still the main criteria companies use to buy servers, and it is how most of us think about servers.

During all the hoopla at LinuxWorld two weeks ago, buried deep inside a heap of press releases, was a statement from IBM that it had 4,600 commercial Linux customers. I've been mulling that number, and I have come to the conclusion that it is pretty pathetic. In early 2000, IBM quietly and secretly caught the Linux bug. What we didn't know at the time was that the company committed $1 billion to developing Linux support for its hardware, software, and services, and in a December 2000 LinuxWorld keynote Gerstner said IBM would spend another $1 billion in 2001. So IBM has spent at least $2 billion on Linux--and very likely more, considering that we are three quarters of the way through 2002--and all that Big Blue has to show for it is 4,600 customers who bought hardware, software, and services for commercial Linux applications? Now that is a good gauge of IBM's commitment to Linux. This can't be a profitable venture for IBM, at least in the short term, and in this regard IBM is no different from every other commercial Linux vendor trying to make a buck on all the hype.

I'm not saying that IBM hasn't spent its money foolishly on Linux. Vendors follow the hype as much as they create it, and it gets hard to separate enthusiasm based on talk from opportunity based on reality. But maybe IBM is making a long-term bet on Linux, which is but one open source product, when what it should have done five years ago was make a smarter and longer-term bet on open source in general. It may be too late to change history, but it may not be too late to take OS/400 open-source and to thereby make the iSeries platform part of the vanguard, instead of the old guard. I know there are legal and intellectual property issues lurking in any strategy that might make OS/400 open-source. But let's look at the possibilities.

Here are some good reasons (in no particular order) why IBM might want to think about creating Open Source/400.

• By making OS/400 open-source, it would be possible for OS/400 to be ported to other hardware platforms. IBM has married OS/400 to its PowerPC and Power4 platforms, but, as we all know, OS/400 rides on top of a hardware abstraction layer that allows it to move to other hardware relatively easily. "Relatively" doesn't mean that this process is easy. It is not. But now that the bulk of OS/400 and its microcode are written in C and Java, porting would be a lot easier than IBM's jump from 48-bit CISC to 64-bit PowerPC RISC chips in the mid-1990s. It might even be possible to backcast OS/400 V5 to CISC-based AS/400 servers, for all we know. Even if it is not economically feasible for IBM to do this work, it might be something the OS/400 community would be willing to do for itself.
• Open source OS/400 could very likely lower OS/400 development costs for IBM, while increasing the features and functions it could deliver. Development projects that IBMers have been itching to do, but can't find funding for, could be farmed out to college students and other open source developers.
• Open source OS/400 would attract the interest of college students and university professors, who generally don't get to look at the guts of a commercial-grade, industrial-strength proprietary operating system. The reason why Linus Torvalds created Linux is that he thought other operating systems available at the time sucked, and that he could do a better job himself. The reason Linux took off is that untold thousands of student and professional developers have been involved in creating and extending the modest Linux operating system. Millions more play a more modest but no less important part, by playing with the Linux code these programmers create.
• Open source OS/400 would seek out and make use of the substantial programming skills of the million or so AS/400 and iSeries programmers out there in the world.
• Open source OS/400 would make OS/400 customers proud of what they have invested in, and make IBMers proud of what they created. The core programmers for System/3X, AS/400, and iSeries operating systems and middleware over the past three decades have not been given proper credit for their brilliance. I can name only a handful of these people. The Linux culture treats its core programmers and stewards with proper respect and can rattle off their names and accomplishments like members of their favorite sports team.
• Open source OS/400 would help Linux to become more credible. Let's face it; Linux needs some serious help when it comes to scalability. It runs well enough on four-way servers, and is said to scale to eight-processor machines. The AS/400 was there five years ago.
• IBM says that OS/400 is bundled for "free" on the iSeries, so, technically, there is no revenue stream to protect. (I believe this statement from IBM is a bunch of nonsense, and I believe OS/400 sales and profits are tracked by Software Group and counted in its quarterly and annual reports.)
• Last but not least, open source OS/400 would be fun and invigorating for customers, developers, partners, and even Big Blue.

There are plenty of obstacles to creating this so-called Open Source/400. If Linux coders got a look at the guts of OS/400, for instance, they would undoubtedly steal the ideas. IBM might not want the core code behind DB2/400 to be out there for the world to see. But if IBM wants to increase its services business in the midrange, and wants to drive server hardware sales in the midrange, then making OS/400 open source might be a good way to accomplish both of these goals. Let me know what you think, at tpm@itjungle.com.

Sponsored By WORKSRIGHT SOFTWARE

On June 30, 2002, $$$$$$$$    Postal Rates went UP!    $$$$$$$$ On July 1, 2002, $$$$$    you wanted your postage bill to go down.    $$$$$ We have the solution! CASS certify your mailing names and addresses and presort your outgoing mail and save. Our CASS certification software ensures that your address files have valid ZIP Code and address information. Our presort software ensures that you can properly prepare you mail for delivery to your Post Office. WorksRight Software, Inc. is the number-one source for iSeries and AS/400 CASS, presort, ZIP Code, and area code software and data. Visit our Web site - www.worksright.com - to learn more about our CASS and presorting software, or contact WorksRight Software, Inc., phone 601-856-8337, e-mail software@worksright.com.