Abstract-Service-oriented architecture has become the standard paradigm for software component integration. However, with the permanently increasing amount of available services and dynamic changes, the complexity of such service infrastructures, their maintenance, and consequently the expenditures spent for their operation increase equally. To deal with these effects, an improvement of service composition and discovery becomes necessary, especially a higher degree of automation. Following the idea of Autonomic Computing, which similarly aims at automating processes and workflows to a high degree, service composition and discovery have to proceed autonomously, which will on the one hand side reduce human involvement to a minimum, but on the other side require certain capabilities on the part of these mechanisms. For these purposes, in this paper we define prime criteria that have to be fulfilled for an autonomic service discovery. Based on that we present a comprehensive survey on existing service discovery approaches and evaluate to which extent they already fulfill these criteria. As a result, the paper reveals that there already exist some approaches that support or even fulfill a couple of the proposed criteria, which principally enables autonomic properties, but what is missing is an holistic approach focusing explicitly on providing autonomic properties.