Component-based, service-oriented digital library architectures are being used to provide superimposed information services such as annotations. Although much attention has been paid to the issues in building components for these services, not enough attention has been paid to their deployment—specifically to distribution. We believe that matching the location of executable and data components to the needs of patrons and digital libraries can improve the overall system performance. We describe five distribution alternatives for providing superimposed information services in a digital library and discuss the trade-offs for each alternative. We define some metrics to compare the performance of the alternatives, and use the metrics in a qualitative evaluation of the alternatives. We also discuss potential barriers for performance and means of improving performance. We use our middleware architec ture for superimposed information management, called the Superimposed Pluggable Architecture for Contexts and Excerpts (SPARCE), for illustration.