TCS Daily

Every Market That Rises Must Converge

By Chris Nosko - April 1, 2005 12:00 AM

The advent of open-source software has been hailed as the most significant event in computing since Apple played David to IBM's Goliath. Yet, while open-source code can be found practically everywhere these days, the companies dedicated to bringing it into the mainstream have had a rough time staying afloat.

To economists, open-source's growing pains offer fascinating insights into the impact of technological change on the way markets are organized. But to firms now investing billions of dollars in IT infrastructure, there's nothing academic about the issue. Since computer software is such an important part of the production process in so many different industries, choosing the right software model (in terms of development, licensing, and distribution) can make a significant difference for the bottom line.

Traditional software companies generate revenue by licensing their intellectual property -- a strategy usually thought not to be feasible for open-source software firms because they retain only limited rights in their inventions. During the frenzy of the high-tech boom, the question of how to capture the fruits of innovation without owning the tree was largely ignored. Indeed, half a dozen companies waving the open-source banner managed successful IPO's, even as they rejected traditional intellectual property protection offered by commercial licensing. Their nominal game plan: sell support services to open-source software adopters.

But when the dust settled, these companies were forced to rethink (or more precisely, to think for the first time about) ways to generate revenue. Which was easier said than done: Since most open-source enthusiasts identified more with community than with property, charging for what had been free ran counter to their beliefs. Besides, joining what many viewed as the Dark Side didn't necessarily make good business sense, since they relied heavily on the open-source community for good will as well as for free help in debugging and improving code.

The approach that some of these companies have adopted is to honor aspects of the open-source model while inching toward reliance on fees for the use of intellectual property. To this end, firms have created hybrids combining elements of free code with code that is licensed for cash. Open-source companies MySQL and Red Hat typify this creep toward middle ground.

Ironically, firms known for licensing software in traditional ways are hoping they can become more profitable by adopting aspects of the open-source software model. Thus Sun, heretofore a symbol of capitalism unbound, has opened its Solaris operating system in hopes that wider use will stimulate demand for Sun hardware and consulting services. For its part, Apple has incorporated a large portion of one open-source operating system called FreeBSD into its widely admired OS X operating system and opened the core of this operating system to public use. Even mighty Microsoft has adopted some aspects of the open-source philosophy through its Shared Source Initiative, which allows some developers limited access to code from some Microsoft products.

Understanding Open Source

At one end of the open / proprietary spectrum stands the Free Software Foundation (FSF), headed by veteran programmer Richard Stallman. Stallman argues that society is best served when software is freely shared. Market economics, he says, has no legitimate place in software development.

The Foundation has long aimed to displace proprietary software by developing free versions of popular applications. Its first project was a free operating system based on the Unix operating system. The effort was eventually combined with code written by the legendary Linus Torvalds to form the widely used Linux operating system.

To the Foundation, allowing users to do whatever they want with free code just won't do -- some, after all, might choose to profit from it by incorporating it in commercial products. Instead, the Foundation uses copyright law to keep them on the straight and narrow path. Under the Foundation's General Public License, software derived in the smallest part from free code must remain free.

A more moderate vision of open-source came from a partnership between software developers at AT&T and researchers at Berkeley, who spawned a variant of Unix. The Berkeley developers were more interested in getting credit for their work than in earning royalties. To this end, they created a license that required that contributors be acknowledged, but allowed derivative software to be distributed under proprietary licenses. Accordingly, Apple could take code licensed under the Berkeley Software Distribution, modify it and sell the hybrid as a new operating system for the Mac, protecting the source code against unlicensed use by not publishing it.

Open Source at the Crossroads

A relatively new business approach to software retains elements of the open source model, but generates revenue through the clever use of intellectual property. The product is distributed with open code, but business users must purchase a license to use it. By contrast, non-commercial users don't have to pay, giving the product the benefit of feedback from a larger network of users and in the process making the commercial version of the software more valuable to business.

MySQL, the world's most popular open source database system (with over five million active installations), is a fine example. A rudimentary version was released on the Internet in 1996, but it has since evolved into a product that goes head-to-head with Oracle Database, IBM DB2, and Microsoft SQL Server. In 2001 the principal developers formed MySQL AB as a for-profit entity, with investments from companies including the giant Dutch bank, ABN Amro.

Anyone can download the MySQL source code for free and redistribute it under the FSF's forever-free license. But a different license, for companies planning to customize the product for commercial purposes, removes the restrictions, thereby allowing purchasers to integrate it with proprietary products. Developers can take MySQL, customize it, and sell it as part of a larger product such as a commercial accounting package.

Why go to the trouble of distributing the free version? Database software is almost always customized to meet specific needs. And as more people use a particular package, more developers become trained in customizing the system and more software add-ons are created. This makes the system more valuable to each user, since trained developers and convenient add-ons are easier to find and cheaper to use.

The no-cost licensing option encourages snowballing network effects by creating a pool of developers familiar with the product. Meanwhile, commercial users who intend to embed MySQL into other products are more likely to purchase a commercial license for MySQL if wide use of the free version has made the software easy to maintain.

Unlike MySQL, which started with a mixed licensing agreement, Red Hat began as a pure open-source company in 1995. The company adds value in three ways. First, it integrates components of the Linux operating system into a cohesive package. Second, it adds its own software to provide a better user experience and to make installation and updating easier. Third, it sells support services and certifies that external administrators and engineers are qualified to work on Red Hat products.

All software that is included in the Red Hat distribution is open-source and all software that Red Hat adds to the package is distributed under the General Public License. But the company uses trademark protection and contractual agreements to lock-in users, and therefore operates a step closer to the proprietary model than is first apparent.

When Red Hat first began selling a version of Linux, it concentrated on generating revenue through sales of a packaged product containing a manual and an installation CD. Hence customers were paying for an integrated physical package that could be had less conveniently at no cost. The only problem: sales didn't generate much revenue.

So, in 2003 the company split Red Hat Linux into two products- -- the Fedora Project, a more traditional open source project, and Red Hat Enterprise Linux (RHEL), the flagship product. Fedora is the place for experiments to run and outside developers to submit code, while RHEL is a stable version of Linux for paying customers.

Along with the split came a new licensing agreement. The RHEL source code is available from Red Hat, but the code that computers actually use to run the operating system is available only with the purchase of a support subscription. And these subscriptions must be purchased for each PC -- just like proprietary software licenses.

One of the most fundamental problems with generating revenue from open source software licensed under the GPL is that anyone can take the source code, "compile" it to make it possible for PCs to read it, and resell it without incurring the original creator's development costs. Red Hat cleverly sidesteps this problem with trademark protection. Anyone could in theory rebuild RHEL from the free source code, but she would have to strip out all references to Red Hat if she wished to distribute it in order to comply with trademark law -- an expensive task.

Red Hat thus has it both ways. It is still distributing Linux, thereby benefiting from all the development and debugging work on Linux done by the open-source community. But trademark protection still allows the company to make a buck.

As with MySQL, what economists call price discrimination is at play. Here, though, the market is segmented according to whether users prefer to pay for the software in terms of time and expertise (downloading and compiling the code themselves) or in cash (subscribing to RHEL).

Convergence From the Proprietary Side

Meanwhile, traditional proprietary software firms, aware of some advantages to the open source development model but wary of giving up intellectual property, have been approaching the same issues from the opposite direction.

Apple, for example, purchased NeXT, a failed computer company started by Apple founder Steve Jobs, along with its operating system NextStep in 1997. Apple was under no obligation to release the source code for any portion of the new MacOS X operating system that it based on NextStep. But in 1999 the company chose to release pieces of it under an open source license. Since this open-source package, collectively dubbed Darwin, does not include critical components usually available in modern operating systems, there was little chance anyone would invest the considerable resources needed to transform it into a fully operational open source product that could compete with the MacOS X.

Still, why did Apple bother? It secured the involvement of the open-source community in the ongoing process of improving and debugging the code on which the Mac OS X is built. And it encouraged outside developers to try experiments with the code that might later broaden the commercial appeal of the operating system, but are not a priority for internal Apple developers. For example, outside developers have focused on creating a highly functional version that is compatible with Intel microprocessors.

Long before Sun Microsystems chose to open the source code of Solaris, its crown jewel operating system, the company was experimenting with less dramatic ways to coexist profitably with the open-source community. In 2000 it released the source code to a suite of office productivity applications that the company obtained when it purchased a German software company. Simultaneously, Sun released a number of commercial versions of the suite -- more polished versions with all the bells and whistles that users have come to expect from commercial software. Sun thus uses the open-source version as a loss leader for the premium product.

Early in the 1990s, Sun developed the programming language called Java, which was designed so that programs can be written just once but run on many different operating systems without alteration. Although it has not fully lived up to the initial hype, Java has succeeded in some niches. And in 1999, the company decided to release the Java source code under a hybrid license called the Sun Community Source License (SCSL).

The SCSL makes the source code available, but does not allow the product to be redistributed in any form without a license purchased from Sun. In essence, release of the source code helps developers to create and test their own products and stimulate interest in the original product, while still generating revenues for Sun from commercially distributed derivative works.

In many ways the SCSL is similar to the MySQL licensing scheme. The economic principles are the same: generate a snowball effect by distributing source code, and profit from the resulting popularity by charging for commercial licenses. But in contrast to MySQL, Sun controls everything that gets put into the product, along with whether derivatives can be distributed.

If the Free Software Foundation inhabits the open end of the software spectrum, Microsoft is typically portrayed as residing at the other extreme. Open-source enthusiasts have long dismissed Microsoft as the author of inelegant, inefficient code, while Microsoft CEO Steve Ballmer once described software released under the Free Software Foundation's license as a "cancer."

Nonetheless, in May 2001, Microsoft announced its Shared Source Initiative, which opened the source code of many of Microsoft's products (including Windows) to some software developers. The licensing scheme is full of restrictions on who can see the code and under what circumstances. Moreover, to preserve its copyright, Microsoft bars the use of code from Shared Source products in other products. Still, the initiative is evidence that Microsoft is moving toward the middle.

Microsoft gains from the better bug-reporting, while developers benefit from an increased ability to debug and improve their own programs that run on Microsoft platforms. Microsoft doesn't have the commercial incentives of, say, Sun, to embrace open source wholeheartedly. But it apparently is ready to search for common ground that preserves its ability to use the intellectual property to generate revenue.

Where Next?

The search for ways to profit from software is leading companies into niches all along the development, licensing, and distribution spectrum. Even users who end up paying for software that was once free are likely to be better off as a result since companies are more likely to support profitable products for longer stretches of time.

By the same token, the fact that software companies that used to rely on proprietary models are adopting at least some aspects of the open-source approach is plainly in the interest of customers. Opening access to commercial source code means faster debugging. It is also likely to improve the quality of software developed on the open-source model, since no amount of copyright or patent protection can prevent programmers from using insights gained from access to protected intellectually property.

The only real downside is that the process of convergence is likely to be messy. But order is the last thing one can expect in a field in which rapid change is the only certainty. The intangible nature of the property embodied in software, mixed with the very tangible consequences of success or failure in the software industry, almost guarantees a wild ride.

Chris Nosko is an independent analyst specializing in the economics of technology. Anne Layne-Farrar and Daniel Garcia Swartz are with LECG, the economics consulting firm. A more detailed version of this paper can be found at


TCS Daily Archives