As the number of devices in a pervasive environment is increased, the number of components available on the network also grows rapidly. In such cases, it is possible to compose various applications through a combination of different sets of components. Considering the multifaceted problem of having varying device capabilities supporting a different set of protocols, and each device hosting a number of components providing the same functionality, it becomes very difficult to choose a particular device hosting a required component which can be the best-fit for the user. This becomes practically impossible when the required components are distributed across various devices in the networked environment. We propose a solution for dynamic user task composition considering user preferences, device capabilities, and heterogeneity of communication protocols. With our proposed approach, a user task can be instantiated in different environments using a different set of devices and components, depending upon their capabilities and user preferences. We propose mechanisms for modeling device capabilities and user preferences and for modeling the user task as a graph. We then propose algorithms for selection of devices based on user preferences and task requirements. Since the underlying network is also modeled as a graph, we describe an algorithm for mapping of services in the user task on to the components distributed across devices in the pervasive environment. We also give an overview of our initial implementation and some results of our evaluations