Moving Toward the Middle: The Case Against the End-to-End Argument in Home Networking.
Kenneth L. Calvert, W. Keith Edwards, Rebecca E. Grinter · 2007
Home broadband adoption is growing rapidly in much of the developed world [23], leading to increasing use of networking inside homes to allow sharing of devices, content and of the Internet connection itself by multiple hosts inside the home. The increasing penetration of home networking enables new products and applications in health care, entertainment and security, as well as in areas currently unforeseen. Unfortunately, all is not well in the connected home, and the set of problems facing users of home networking is becoming a key impediment to the promised benefits of home networking. Consumer statistics illustrate the problem: In 2003, home networking gear was the most returned item at “big box” electronics stores in the US [17]. In 2006, a quarter of wireless access points purchased by consumers were returned [12]— not because the devices “broke”, but because their users were unable to properly set up and integrate them into their home networks; complexity is reliably cited as the key impediment to home networking [11]. Even if deployed successfully, home nets are often misconfigured, posing a threat to the larger Internet as evidenced by the existence of large-scale “botnets.” Why should these problems be the domain of the networking research community? Isn’t this a simply an “implementation problem” that vendors and ISPs will address? We claim that solving the problems associated with home networking is not simply a matter of building a better home router, providing more clear documentation, or better technical support. Rather, many of these problems arise from a mismatch between certain aspects of the present Internet architecture and the characteristics of the home environment. In particular, the endto-end principle, which has shaped many of the design choices of the Internet [16], is a contributing factor to the difficulties seen in the home (a point that has been raised by others [18], perhaps most notably by Blumenthal and Clark [3]). We argue that a more comprehensive system solution is required. In this paper we examine the problems that arise from “bringing the Internet home”—applying the same protocols and architectural principles designed for the Internet-at-large in the home environment. This examination is rooted in a series of studies that have explored the problems of networking from a human perspective. We tie these end-user visible problems to the architectural choices that result from the end-to-end argument, and present a new architecture for the home network that elevates ease of installation and use, evolution, and troubleshooting to the same level of importance as the application neutrality and core network simplicity goals dictated by the end-to-end argument. The rest of this paper is organized as follows. Section II considers why the Internet architecture was brought into the home initially, as well as why it might be inappropriate for the home setting, based on user studies of the travails of networking in the home. Section III presents a set of design goals for any home network architecture. We then present a prototype design and argue that it satisfies the goals. We discuss related work in Section V, and conclude with some thoughts on the implications of our work.