How (not) to design an Internet : principles and properties of network architectures

Abstract

The TCP/IP internet architecture (IA) has been tremendously successful and the Internet has become a dominating component in the world's economic and political landscapes. However, the Internet is based on a prototype network that outgrew itself before its teething troubles were rectified and it has serious issues with flexibility, mobility, scalability, and security: flexibility is lacking due to a bottom-up intensional design, whereby internal identifiers and semantics are exposed through the APIs; mobility is either not transparent to applications and requires DNS involvement, or it does not scale; scalability is sacrificed for mobility or to avoid network renumbering; and security needs to be provided by applications, alternative protocols, or middleboxes. Multiple future lA (FIA) proposals as well as numerous augmentations to the classic TCP/IP IA have been devised to remedy these issues. They have very different goals, designs, and properties, which make them difficult to compare meaningfully. Consequently, the research community seems unable to agree on common solutions to the deficiencies. In this dissertation, we seek to address this situation by providing a qualified assessment of several contemporary and future IAs with a special focus on the classic Internet (IP version 4 (IPv4)-based TCP/IP and DNS), modern Internet (the Internet with its modern augmentations such as IP version 6 (IPv6), Software-Defined Networking (SDN), Mobile IP (MIP), and the Locator/Identifier Separation Protocol (LISP)), the Recursive Internet Architecture (RINA), and Named Data Networking (NDN).