Many architectural elements of today's Internet implicitly assume that hosts remain attached to the network over extended periods of time. However, there is much to be gained by supporting a style of networking that does not require an end system to maintain full, ongoing connectivity in order to maintain its network "presence." In this project we explore a style of networking that we term "selectively-connected", by which we mean an end system, can knowingly manage the extent of its network connectivity in response to internal or exterior events, as it anticipates changes in connectivity.
While selectively-connected networking also has applications for end systems that enter outage periods (e.g., a user closing a notebook, or a mobile device entering no-coverage area), one highly significant form of operation it can enable concerns placing end systems into some degree of "sleep" in order to operate with much greater energy efficiency. Such sleeping not only can benefit portable devices by greatly extending their battery lifetime, but can also realize energy savings at a national scale by enabling desktop systems and set-top devices to enter states of greatly reduced processing without sacrificing their network presence.
In this project the researchers undertake initial designs of new architectural components for better supporting selectively-connected networking, by which sleeping hosts can retain their standing in the network or delegate agents to act on their behalf during their absence. These span: exposing selective connectivity; evolving soft state into notions of "proxyable" or "limbo" state; facilitating host-based control; introducing "assistants" to work in concert with sleeping end systems; exploring primitives that applications might use to express the semantics they wish to preserve when selectively-connected; and considering network links that can themselves sleep when the end systems they serve are likewise sleeping.
