How We Tested the Dynamic Content Tools

Our simulated Web e-commerce sites used either Netscape Enterprise Server 3.5 or Microsoft Internet Information Server (IIS) 4.0. Both Web servers ran on Windows NT 4.0 with Option Pack 3. The server was a Gateway NS-8000 333-MHz dual-Pentium computer with 512 MB of RAM. On the client, we assumed only the presence of a Java-enabled, frames-capable browser.

To measure performance, we timed the loading of Web pages in each product's environment and the processing of each sales transaction. In each case, we rebooted both the Web server and the clients to eliminate the effect of caching on performance. We also examined the size (and thus the download time) of each product's emitted Web pages. Because our test pages incorporated dynamic HTML, simply looking at the sizes of static pages on the Web server wasn't enough. Instead, we used a Network Associates Sniffer to discover the sizes of the pages as they flowed over our test network.

As we tested for performance, we found that a "quick trip to the store" depended on more than just quick-loading Web pages. It involved good design, easy-to-follow navigation and, for a large catalog, the availability of a useful search engine. The speed with which you can turn the pages of a printed mail-order catalog isn't the most critical factor if you're in a hurry. Similarly, an e-commerce Web site can help a rushed shopper more by offering product bookmarks, a bottom-of-the-screen shopping-cart contents reminder and an intuitive flow of Web pages.