Web browsers have become the method of choice for access to bibliographic databases, almost entirely replacing proprietary clients for new end-user systems. Going one step further and programming the client in Java to run within the browser, rather than relying entirely on HTML, opens up a new level of functionality. SecondSearch is an experimental system under development to explore both performance and functionality issues of this approach. It uses a tree structure to display a hierarchical view of records, giving the user the ability to zoom in and out of sections of the selected databases. The system is designed to be scalable to thousands of users browsing and searching hundreds of millions of records. A major challenge has been to create an interface that is simple, powerful, does not place an exorbitant load on the server, and is suitable for dozens of databases, each containing millions of records.