Software-defined networking (SDN) is arguably the single most important development in the networking field in the past few years. SDN decouples how packets are forwarded between routers/switches (data plane) from how packets must be handled (control plane).

Feamster et al. present a history lesson on how different technologies and paradigms played a role in shaping what is now called SDN. A chronological overview of past efforts in the area of programmable networking is depicted from early research on active networks, through control-data separation, and finally on to the more recent OpenFlow and network operating system efforts. Furthermore, the authors discuss what has driven the interest from academia and industry through the years: initially, it was the search for more pragmatism, and later it was the need to generalize previous solutions.

Even experienced researchers and practitioners will appreciate the chronological relationships highlighted by the authors. Not surprisingly, many concerns faced earlier by more primitive solutions surfaced later in the SDN domain, including the need for distributed state management. The reliance on existing protocols and application programming interfaces (APIs) significantly limited what could be done in the early 2000s, but the authors emphasize that present-day SDN and OpenFlow as its primary realization build on similar principles for supporting a more general manner of network programming.

If one thing is missing in this paper, however, it is a discussion on abstractions for expressing network programmability and for the specification of management functions. This would have included early work on policy-based network management (PBNM), the Internet Engineering Task Force (IETF) script management information base (MIB), and more recent efforts along the lines of Pyretic.