In this paper, using RETINA, a real-time navigation system, as an example, we study the important design issues underlying the processing of location-dependent continuous queries, especially those requiring access to data describing the current status of a dynamic environment and possessing spatial properties. To minimize the probability of missing the arrival deadline associated with each navigation request, we use a time-stamp with prediction scheme to model the traffic data and replicated dynamic directed graphs to organize the traffic data required for path searching and path calculation. Correctness of the best path calculations and scalability of the system are improved through an adaptive Push or Pull (APoP) scheme to monitor the best path and traffic data in navigation.