In a distributed metacomputing environment, facilities for the movement of code from one server node to another are highly desirable. Often movement of code to data is preferable to the movement of data to code, particularly where the data is large. In this paper we describe our `Code Server' or database of Java byte code which is searchable on code attributes. We discuss design and implementation issues and results of performance benchmarking. We focus on a practical image processing example. We also discuss how our implementation addresses issues such as namespaces, client-server communication and distribution. Our results show that code serving appears viable, but more work remains to identify a suitable general database schema for describing code attributes.