Ubiquitous User Modeling aims at providing personalized services to inhabitants of smart environments. Current research in ubiquitous user modeling focuses on two directions. The first is a practical approach that tries to resolve current problems of sparseness of data and heterogeneity of user modeling techniques and representations by mediation of user models or building hybrid systems. The second approach is based on semantic standardization of user modeling enabling user modeling data exchange and sharing by using a common user modeling ontology and language. Although both approaches have their limitations, their integration has the potential to leverage their advantages and overcome the limitations. This paper discusses initial work done in this direction, suggests a path for such integration, and points out research directions aimed at bridging the gap between these approaches.