Roger Bowler Responds to IBM Patent Attack on Open Source

by Roger Bowler
Creator of Hercules and Co-founder of TurboHercules

April 06, 2010

As many of you know, the company I founded to promote the Hercules open source mainframe emulator, TurboHercules SAS, has filed an antitrust complaint against IBM with the European Commission in Brussels. We are not asking that IBM be subjected to punishing fines or anything like that. We simply want IBM to agree to allow legitimate paying customers of its z/OS mainframe operating system to deploy that software on the hardware platforms of their choice – including, should they so choose, on low-cost servers using Intel or AMD microprocessors and Hercules.

I want to make clear that we undertook this action reluctantly, and only after a long period of reflection during which we reached out to IBM to see if there was some way to resolve our differences amicably. I regret to report that IBM rebuffed our efforts at conciliation, and even added fuel to the fire by launching accusations against Hercules. I would like to take this opportunity to respond to some of those charges.

First let me deal with IBM’s claim that we at TurboHercules are “not really any different from those who seek to market cheap knock-offs of brand-name clothing or apparel”. To that I would like to say, Hercules is not a fake Gucci handbag. Hercules has never pretended to be an IBM-brand product, and no customer would ever mistake it for such. Hercules is a third-party, open source software-based emulator developed in good faith using IBM’s published documentation of its z/Architecture. I know from personal experience that among the thousands of satisfied users of Hercules over the years more than a few have been IBM employees. There was even a time when IBM thought highly enough of it to publish a friendly and quite detailed chapter in one of its famous Redbooks explaining how to use Hercules. This Redbook was freely distributed in print and on IBM’s web site for several years, until one day the Hercules chapter was silently deleted. Since that time of course IBM has launched its own mainframe emulator (zPDT) in competition with Hercules. Readers may draw their own conclusions as to whose software is the “knock-off” and whose is the real thing.

IBM says that TurboHercules seeks a free ride on IBM’s “massive investments in the mainframe by marketing systems that attempt to mimic the functionality” of its machines. By the same logic, one might very well charge that Linux is nothing more than an attempt to “mimic” the functionality of UNIX. Suppliers of UNIX systems such as IBM, HP and Sun have undoubtedly invested large sums over many years in developing these platforms. Is it therefore wrong for Red Hat and SUSE to offer users a less expensive “UNIX -like” operating system that has the conspicuous advantage of being open source and of being able to run on just about any server the user pleases? IBM have found Linux a very useful tool for pursuing their commercial aims, and I do not think that IBM would call Linux a “cheap knock-off”. Both Linux and Hercules are examples of the kind of low-cost innovation that outsiders can bring to the IT industry.

It is very strange for IBM to insist that others no longer have the right to emulate its mainframe instruction set. The entire IT industry is built on the premise that every participant has the right to interoperate on an equal footing with the interfaces published by other participants. For 25 years IBM competed vigorously in the mainframe market with the likes of Amdahl, Fujitsu and Hitachi, and it did not accuse them of “mimicking” its architecture, but on the contrary gladly sold its sophisticated and very expensive mainframe software to any and all customers who purchased “plug compatible” mainframes. The plug compatible industry was based on the fact that IBM’s hardware interfaces were open. It was up to each company to design its own implementation, and competition was fierce. In those days IBM didn’t object to the principle of free competition and open interfaces, and this was a good thing for customers.

Likewise, it would never have occurred to me to attempt a software emulation of the IBM mainframe if IBM itself did not publish thousands of pages of very detailed documentation explaining precisely how to do that. And TurboHercules is far from being the only company to offer emulation software as an alternative to certain kinds of hardware. As a service to its customers, Apple includes a piece of software known as Rosetta with every Macintosh that enables its current Intel-based machines to emulate its older machines based on IBM’s PowerPC. I understand that IBM uses this very same emulation software, which it acquired from a company called Transitive, to allow owners of Linux applications compiled for Intel processors to run these binaries unchanged on IBM Power servers. If IBM itself considers emulation legitimate in its own products, am I not entitled to conclude that what’s good for the goose is good for the gander?

But I don’t wish to portray this issue as merely a dispute between my tiny company and IBM. Much more is at stake. Let us remember that IBM is not the only stakeholder in the vast ecosystem that has grown up around its mainframes. This ecosystem not only earns IBM well-deserved profits on its past investments, but also provides the infrastructure for many of the world’s most important business applications and – last but not least – employment for tens of thousands of highly skilled people, most of whom do not work for IBM. From the very beginning of the Hercules project, it has been my desire to serve this broader mainframe ecosystem. Tiny TurboHercules is not going to bring down mighty IBM or the mainframe technology of which it is rightfully proud. But we are stirring things up a little with innovations that IBM itself did not see fit to bring to the market. Now with the launch of zPDT even IBM has recognized the value of being able to run the mainframe software stack on common Intel hardware. Who can deny that this is an important step forward, one that will certainly serve to enlarge the mainframe ecosystem rather than diminish it?

What TurboHercules offers today is an innovative and very low-cost Disaster Recovery option for mainframe sites that cannot afford a second mainframe or a third party service. This is a customer need that IBM for whatever reason has chosen not to serve. Naturally we hope to go beyond this solution one day to address other unmet needs of mainframe users. For example, some users might find it useful to run their z/OS applications with VMware on the same Intel servers used by their Linux and Windows applications. Hercules can make this possible. In these times when the words “we’re trying to get off the mainframe” are heard all too often, Hercules will also give users a reason to stay on the mainframe and to buy more of IBM’s mainframe operating systems, middleware and other software. But it can only do this if IBM accepts that users who pay for its software ought to have the right to choose what hardware they want to run it on.

IBM complains in its press statement in response to our antitrust filing that we “seek to use governmental intervention” to advance our interests. However, we acted only as a matter of last resort. Indeed, just a few days before we filed the complaint with the European Commission, Mr. Mark Anzani, the CTO of IBM’s mainframe division, wrote to me to allege that the open source Hercules emulator may have violated no fewer than 173 of IBM’s patents or patent applications. Not only is this accusation untrue, but it flagrantly violates IBM’s many past promises not to use its patents as a weapon against open source. At least two of the patents on Mr. Anzani’s list even figure on IBM’s famous pledge of non-assertion of 500 of its patents against open source issued in 2005. Advancing our interests? I see it as defending our right to exist against a company that wields the world’s largest patent portfolio as a weapon against competitors thousands of times smaller than itself.

IBM also accuses TurboHercules of cooperating with Microsoft. Bearing in mind that Hercules works very well indeed on both Linux and Windows, not to mention the Macintosh, we are indeed quite happy to cooperate with Microsoft, Hewlett-Packard, Unisys, Dell, Intel, AMD or anyone else who wants to work with us. We would be delighted to cooperate with IBM, if offered the chance. For example, since Hercules is written in C, it would be quite simple to port it to the Power platform, where it would offer a very cost-effective way of running z/OS together with AIX. I’d bet that more than a few IBM customers would be interested in doing this.

Finally, IBM says in its reply to our filing that “the mainframe is a small niche in the overall server market”. I can’t imagine why the IBM mainframe division allowed their publicists to say such a thing. I am very sure they don’t believe it, and neither do I. On the contrary, I believe that, popular wisdom notwithstanding, the vast IBM mainframe ecosystem has many prosperous years ahead of it, and I am confident that TurboHercules will succeed by serving the interests of all stakeholders in this system.