A new transport encapsulation for middlebox cooperation

Large-scale pervasive surveillance of Internet traffic has led to a rapidly expanding deployment of encryption. Network operators increasingly rely on the use of in-network functionality provided by middleboxes to make their networks manageable and economically viable. Emerging applications such as real-time video conferencing need further support that the transport layer, having become ossified between narrow interfaces and an inability to universally deploy new transports or new options, is unable to provide. To address the collision of these three forces, we present in this paper a new encapsulation and signalling protocol, called the Substrate Protocol for User Datagrams (SPUD), that provides a shim layer for middlebox cooperation. While the protocol itself must be standardized to allow for successful deployment and cooperation, in this paper we additionally assess research questions on how this new protocol can be used to expose information to middleboxes to enable low latency support for new services. We describe our approach based on a prototype implementation available to the community, and show through a small initial measurement study that the use of UDP as a multiplexing encapsulation for SPUD, which allows for NAT traversal as well as rapid deployment in userspace, is a valid strategy in the present Internet. Finally we discuss open issues and research challenges in SPUD's development as an input for both, the standardization community as well as for academic researchers.