Infrastructure microservices such as Service Discovery and Routing need mechanisms to manage distributed state. The most common answer for this is the use of a central, consistent key value store such as Consul, Etcd or Zookeeper. These systems use consensus-based algorithms, such as Raft or Paxos, to provide consistency and failure tolerance. I believe this is a dangerous direction for our industry, and instead we should be focusing on ease of use and reliabiligy. As such, Weave wants its infrastructure microservices to be decentralized, yet easy to install and run; our approach is inspired by the Internet, which is distributed and operates with no consensus. In this talk, I'll explain how we designed our Service Discovery and Address Management using Convergent Replicated Data Types (CRDTs) and Gossip, review the pros and cons of this concept, and how it compares to alternatives.