UC Santa Cruz

Most existing reliable transport protocols in the current Internet architecture, like TCP, rely on an end-to-end connection as the main architectural construct to implement reliable transfer of data. This has become a big problem for content- oriented Internet applications because connections are brittle; in most protocols, the context exchange must be restarted if a connection is lost, a specific site must be selected to start the connection supporting content retrieval, and in-network caching cannot be used without relying on third parties applications such as CDN to replicate content-delivery functionality to only a subset of its users. As a result, several Information-Centric Networking (ICN) architectures have been developed that focus on information-centricity and eliminate connections between remote processes at the transport layer through modifications in the network infrastructure. However, this can only be achieved by requiring a far more complex network layer. In addition, transparent in-network caching is meant to be a vital benefit of these solutions, with the content being cached opportunistically throughout the network where it is needed. However, intermediate routers and caches have access to the content object being transferred to consumers, which poses privacy concerns.

To address these problems, this thesis proposes a new way of improving the Internet’s architecture at the transport layer through a novel method of transport- ing data without using end-to-end connections and enables the use of transparent caching at the transport layer while preventing caches from decoding cached con- tent.

The Internet Transport Protocol (ITP) is introduced as an alternative to the Transmission Control Protocol (TCP) for reliable end-to-end transport services in the IP Internet. The design of ITP is based on Walden’s early work on host-host protocols, and the use of receiver-driven Interests and manifests advocated in several information-centric networking architectures. The design of ITP proves that connections or significant changes to the routing infrastructure are not needed to provide efficient and reliable data exchange between two remote processes on the Internet. Through minor extensions to the Domain Name System (DNS), ITP design can be leveraged to provide efficient content delivery by name over the existing IP Internet allowing it to achieve NDN advantages without the need for a new routing infrastructure. The use of manifests to describe contents glob- ally at the transport layer eliminates the need for a new ICN like technology by allowing data to be cached at the transport layer and preventing these caches from understanding cached contents by enforcing decoding methods as part of the manifests.