The End of Joint Development

Microsoft's success strained its relationship with Sybase. The competitive landscape of late 1993 was quite different from that of 1987 when Microsoft and Sybase had inked their deal. By 1993, Sybase was a successful software company, by most accounts second only to Oracle in the DBMS market. The credibility and visibility Microsoft brought Sybase was far less important in 1993 than it had been to the upstart company of 1987. Similarly, Microsoft had grown a great deal since 1987. The growth wasn't just in revenues (although that was one of the great success stories of the industry), but in an emphasis on enterprise applications, such as Microsoft SQL Server, that Fortune 1000 companies could use as a platform upon which to run their businesses.

The SQL Server development team had grown as well, from a handful of people in 1990 to more than 50 professionals (not including marketing, support, or field operations), with significant additional growth planned. The first-rate team of engineers knew database and transaction processing and the inner workings of SQL Server, and they were experts in developing for Windows NT. Microsoft now had the talent, size , motivation, and mandate , yet was still constrained in what it could do with the product: the 1987 agreement with Sybase had merely licensed to Microsoft the rights to the Sybase product. Because of this restricted agreement, Microsoft couldn't unilaterally implement new features or changes without Sybase's approval. Contrary to what many people thought, Microsoft had no ownership stake in Sybase and by no means could Microsoft simply call the shots.

Obviously, Sybase had different business needs and therefore different priorities than Microsoft. The development team at Microsoft might, for example, want to integrate SQL Server with messaging by using MAPI (messaging API), but because this feature was specific to the Microsoft operating system, Sybase wouldn't be excited about it. As is always the case in development, many features could be done for every feature that will actually be done: features specific to Windows NT didn't tend to interest Sybase as much as those that would benefit its UNIX products.

Sybase engineers had to confront the issue of portability to multiple operating systems. In fact, Microsoft's implementation of version 4.2 for Windows NT was already causing friction because Sybase was progressing with System 10 for Windows NT. Sybase was understandably implementing System 10 in a more portable manner than Microsoft had done. This was entirely rational for Sybase's objectives, but from Microsoft's perspective, it meant a looser bond with Windows NT. System 10 would not, could not, perform as well on Windows NT as the product that had been designed and written exclusively for Windows NT.

Because of the economics involved, as well as the changing competitive landscape, the Microsoft/Sybase agreement of 1987 was no longer working. Microsoft SQL Server was now a viable competitive alternative to Sybase SQL Server running on UNIX, Novell NetWare, and VMS. Far from seeding the market for Sybase, Microsoft SQL Server was now taking sales away from Sybase. Instead of choosing a UNIX solution, customers could buy Microsoft SQL Server at a fraction of the cost of a UNIX solution, run it on less expensive PC hardware, and install and administer it more easily. Although Sybase earned royalties on sales of Microsoft SQL Server, this amounted to a small fraction of the revenue Sybase would have received if the customer bought Sybase products for UNIX in the first place. Microsoft and Sybase were now often vigorously competing for the same customers. Both companies recognized that a fundamental change in the relationship was needed.

On April 12, 1994, Microsoft and Sybase announced an end to joint development. Each company decided to separately develop its own SQL Server products. Microsoft was free to evolve and change Microsoft SQL Server. Sybase decided to bring its System 10 products (and subsequent versions) to Windows NTthe first time that the Sybase-labeled SQL Server would be available for a Microsoft operating system. (The original agreement gave Microsoft exclusive rights on Microsoft operating systems.) Both companies' products would be backward-compatible for applications that had been developed for the shipping version of Microsoft SQL Server. However, the products would diverge in the future and would have different feature sets and design points. Sybase's product would be fully compatible with its UNIX versions. Microsoft's product would continue to be integrated with Windows NT as much possible. In short, the products would directly compete .

SQL Server Performance: No Secret Formulas

Although Microsoft SQL Server is designed for and optimized for Windows NT, it uses only publicly documented interfaces. From time to time, articles or speakers suggest that Microsoft SQL Server uses private, undocumented APIs in Windows NT to achieve its performance. But this assumption is false, without exception. SQL Server's performance results from using the available and published Windows NT services without any compromise. This is something that other products could also do if developers were willing to dedicate themselves to doing the best possible job for Windows NT, without making compromises for portability to other operating systems.

NOTE

---

Although the Sybase product competed with Microsoft SQL Server, Microsoft encouraged and provided support to get Sybase System 10 shipping on Windows NT as soon as possible, because it was important to the acceptance and success of Windows NT. Such collaboration is typical of many other relationships that are competitive on one level and cooperative on another.