May 2000

The Open-Source movement has caused a lot of people to wake up and re-examine the fundamental assumptions of the software business. Some of these long-held assumptions now being challenged are the idea that you sell the software for profit and give away the support for free (or for a small charge); the notion that the greatest thing a company should fear is that somebody will find out how their products really work deep down inside; and of course the belief that operating systems can and should be huge, tightly-guarded monoliths that dictate who can or cannot interface to them.

While IBM has broadly embraced the poster child of Open Source -- namely, Linux -- it has not been as firmly supportive of 100% Open-Source when it comes to its own operating systems. There are several reasons for this reluctance. For one thing, IBM's operating systems are the result of decades of investment, research, and experience with large systems. Also, IBM has spent the last thirty years incorporating many copyrighted and patented features into their own OS's. It would be unreasonable for them to spend millions of dollars not just in development, but also in legal expenses to seek and protect government sanction for these innovations, and then to freely open the vaults to all comers.

Most importantly, IBM has shareholders to answer to. Until the pure Open-Source model becomes a stable, long-term profit center for many companies, IBM will have quite a challenge in convincing shareholders that giving away the "crown jewels" is the right thing to do. Small companies and individual innovators like Linus Torvalds have a great deal more freedom to choose their software paradigm -- open or closed. The stakes are not nearly as high, and they have a greater degree of independence.

However, there are some signs that IBM may be moving toward the Open Source model itself, even with its own products. IBM appears to be taking a "layers of the onion" approach, offering full source for file systems, protocols, and other connectives. Later, IBM will likely open up some of its more common APIs and other features within the software products themselves. Kernel-level code will probably be opened last, if at all.

(For some examples of IBM products that are now Open-Source, see the list of available code at IBM's developer page. The current list is found at

One product that has been the topic of much speculation relative to Open Source is OS/2 Warp itself. The main obstacle to opening the source for Warp is that IBM and Microsoft have joint ownership and mutual contractual obligations on some parts of the code. This is one reason why IBM cannot legally divulge all the wonderful, cutting-edge goodies on the inside that make OS/2 the technological marvel that it is. Microsoft would never allow the software development world to find out just what they have been missing all these years! However, there are financial reasons as well.

IBM sells anywhere from a quarter-million to five million copies of the OS/2 client per year, depending on whom you believe, at prices ranging from perhaps $75 a pop for large orders, to $250 a box for individual full-retail copies. (Surely only government agencies pay that price!!) That's anywhere from $50 million to $1 billion in annual revenue, just from the OS itself. Naturally, far more is made from the related custom applications, services, and associated hardware purchases. (Plus sales of Warp Server, which code IBM will probably always keep in-house.) But it would be hard to believe even a wealthy company like IBM could afford to leave that kind of money on the table. Unless....

I wonder just how successful a "pay for the source" contract would work. Being Open Source only means "zero cost" if people copy the source code and redistribute it. It is possible that in IBM's relatively controlled-environment customer base, an Open-Source-for-Money environment may yield the bulk of current Warp client receipts. The question then becomes, What good is Open Source if only a limited number of people have access to the code?

The result of such an arrangement really would not be "open" in the classic Open Source sense. It would be like Microsoft's fake "openness" in licensing NT source code to a handful of carefully watched buddy-buddy companies. We know just how poorly that has worked in terms of innovation and bug fixes for that platform. Besides, one important reason for having a zero-cost product is to remove most or all of the implied "risk" involved wtih investigating a new, non-Microsoft platform.

Realistically, the only way IBM would open-source the entire Warp client would thus have to address two issues: first, how to remove the Microsoft ball-and-chain, and second, how to ensure sufficient receipts to nullify lost revenue due to unlimited code redistribution. It is quite possible that the first issue is already being addressed. IBM has rewritten the Warp 4 kernel in the latest Fixpack 13 as well as the upcoming Convenience Packs. What if the cumulative rewrites over the past four years have finally eliminated the co-owned code? Then Microsoft would have to keep its mouth shut and watch IBM open a "second front" in the war on Windows. With Linux attacking from the Server and OS/2 freely distributed across the desktop, Microsoft would be stuck fighting an expensive two-front war -- against two products that it could not undercut on price.

The second obstacle, then, would need to be overcome: how to achieve continued revenue from the Warp client without limiting its distribution. The ideal solution would be to form close alliances with several OS/2 ISVs and/or some second-tier or third-tier PC makers, outsource the source code for the Warp client to them for an annual fee, and let them each produce their own Warp client distribution. But there has been no sign of any sort of movement in this direction by IBM, except in having companies like Acer build the IBM PCs.

OS/2 is having a revival, however, and it would be a good idea to watch and listen carefully for any signs of Open-Source strategies filtering into the Warp product group at IBM. It is quite possible that IBM has already achieved a truly Microsoft-free base OS to go along with the Windows 3.X source code they have apparently already acquired from Microsoft. That could mean IBM would finally have the green light to use the full power of Open Source in the continuing battle for OS mindshare. Maybe five years from now, we can say that "OS/2" stands for "Open Source 2"!

Most recent revision: May 31, 2000
Copyright © 2000, Tom Nadeau
All Rights Reserved.