Knowledge-based networking involves the forwarding of messages across a network based on semantics of the data and associated metadata of the message content. However such systems typically assume a common semantic model underpinning the routing which limits their ability to cope with heterogeneity. In contrast, the authors have developed a semantic-based publish/subscribe system that is unique in allowing several semantic models to support routing. This paper examines the content heterogeneity problem in a Knowledge-based Network (KBN) implementation and evaluates a mechanism for efficiently and dynamically loading ontological mappings for use with distributed and heterogeneous knowledge-based applications. It compares a number of strategies that use pre-existing semantic mapping information stored in KBN routers. Evaluation results show that this mechanism can effectively solve the heterogeneity problem.