Pricing models for Microsoft server products have evolved somewhat "organically" over the years, with each product group introducing its own peculiarities to meet the needs of its particular marketplace. As a result, pricing has become an exceedingly complex discussion. It is challenging for a customer to ascertain the initial set of purchases necessary for license compliance, no less keep track of compliance on an ongoing basis. Even within Microsoft there is confusion—the most common type of question posed by the field sales force to the server product groups concerns the topic of pricing.

Starting in 1994, Microsoft began charging a set fee for the right to install one of its server applications on a server machine and required the purchase of a Client Access License (CAL) for each desktop PC used to access the server. This is known as the Per-Seat pricing model. The company later offered a second pricing option—known as Concurrent-Use (also called "Per Server")—for a subset of its server products. In the Concurrent-Use model, the CALs are applied to a particular server machine and determine the maximum number of simultaneous clients that can connect to that specific server. Concurrent-Use is more cost-effective than Per-Seat when a server application has only limited use within an organization. For example, when a large number of desktops access a single server application infrequently, the purchase of a CAL for each workstation (under the Per-Seat plan) would make the server application a prohibitively expensive solution and thus the Concurrent-Use option is preferable.

In late 1996 Microsoft was compelled to add a third pricing scheme. As Microsoft server products began to be used in an Internet capacity—where it is difficult to predict the number of simultaneous users in advance, or to license each user’s desktop—the company came up with a new special-purpose pricing model, called an Internet Connector. In this model, the customer pays the standard fee for the right to install the server application on a server machine, plus a fee for an Internet Connector license that entitles an unlimited number of users to access the particular server via the Internet. An Internet Connector is like purchasing an infinite number of CALs under the Concurrent-Use model.

SQL Server was the first product to offer an Internet Connector option. The first incarnation of SQL Server’s Internet Connector license allowed an unlimited number of client devices to connect to a SQL Server machine as long as those connections were provided "through a Web Server." Microsoft subsequently modified the licensing terms to limit access to the purchasing organization’s customers only—the organization’s employees and employees at the organization’s suppliers were excluded from accessing the server via the Internet Connector.

By 1999, Microsoft had over a half dozen server-based applications, each supporting between one and three different pricing models and each having its own special set of pricing "footnotes." For example, SNA Server 4.0 required the purchase of CALs only for a certain subset of connectivity services, but not for others; SQL Server 7.0 required an Internet Connector license for each CPU in a Web server machine, whereas Site Server 3.0 Commerce Edition did not. Most customers were left scratching their heads. Often unable to calculate licensing costs with precision, companies simply made a "best effort" computation and wrote a check to Microsoft. Microsoft realized the confusion was costing them revenue and that change was necessary. Especially hard hit were sales of CALs, which Microsoft suspected were selling at a rate much less than license compliance would dictate.

Besides customer confusion and lost revenue, Microsoft saw another impetus to update and simplify server application pricing—its evolving Next Generation Windows Services (NGWS), now called .NET ("dot net"), initiative. The .NET vision—in which multiple Web sites/services interoperate seamlessly behind the scenes to deliver a consumer experience—demanded a new pricing model. To be easy to comprehend and practical to apply, the new model couldn’t set limits on the number of simultaneous users or connections, couldn’t require Internet access devices (clients) to be specially licensed, couldn’t draw a distinction between use inside or outside the corporate firewall, and couldn’t distinguish between use by the purchaser’s own employees and suppliers vs. the purchaser’s customers. And yet, the new model had to be a reasonably good measure of actual use so that customers would perceive it as "fair." A Per-Processor pricing scheme—already used by several competitors—met these requirements, and Microsoft chose to roll out Per-Processor pricing in conjunction with its launch of the Windows DNA 2000 server product line.

Back to associated article: Some Windows 2000 Server Products Get Per-Processor Pricing