As the most important component of pervasive computing, context-aware system gets more and more attention from the researchers. One of the basic technologies used to implement the contexts' description and reasoning in context-aware systems is ontology which has the big advantage at knowledge presenting and sharing. However, the shortage of the existent methods with ontology is obvious that most of them are predefined which makes them too fixed to provide the flexible services. In this paper, a Distributed Active Service Providing System (DASPS) is proposed to provide more accurate and more proper services to the end users in an active manner through ontology based contexts' similarity comparison. Moreover, a semantic-adaptive service choosing machine (Sasrc_Ma) is used in DASPS to improve the accuracy of the contexts' similarity which can dynamically adjust its computing result to satisfy the real situation of the users by considering the forwarding feedbacks of users. An effective negotiating mechanism is used to solve the conflicts occurred in this distributed application when dealing with the concurrent transactions.