Distributed systems are becoming increasingly complex. We are moving from a world where we provide domain-specic middleware platforms (e.g., for Enterprise systems, Grid, MANET, ubiquitous environments) to one where these technology-dependent islands are themselves dynamically composed and connected together to create richer, dynamically deployed systems. Existing middleware approaches and paradigms are simply unable to cope with the demands of such heterogeneous and dynamic environments. Indeed, as we move towards a world of systems of systems, we can say that middleware is in crisis, being unable to deliver on its most central promises, which is oering interoperability, i.e., the ability of one or more systems to connect, understand and exchange data with one another.We argue that this requires a fundamental re-think of the architectural principles and techniques underpinning middleware platforms. We need to turn from relatively static solutions based on promoting a particular interoperability solution or bridging strategy, to much more dynamic solutions where we generate appropriate machinery for interoperability on the fly. This promotes an approach that may be termed emergent middleware designed to solve interoperability at runtime according to what is discovered and needed in a given context. This paper reports on the results of the Connect project, a collaborative initiative bringing together experts in middleware and software engineering, semantic modeling of services, and formal foundations of distributed systems, which together provide key building blocks for enabling emergent middleware.