Computer networking is notoriously riddled with protocol acronyms, header formats, marketing terms, and other domain details. There is little evidence of modularity or separation of concerns. Every new technology engenders a wave of research papers, amply demonstrating that there is scant understanding of principles that underlie all networking technologies.
This situation makes networks far too difficult to design and manage. It also makes networked applications far too difficult to build and maintain, as applications are forced to work around the quirks and deficiencies of the underlying communication services. The situation impedes evolution toward an Internet that satisfies the demand for an ever-wider range of devices, applications, and resource policies.
To put the discipline of networking on a stronger foundation, and guide future research toward network services that fully satisfy their requirements, we need to identify useful abstractions of the purposes, functions, and mechanisms of networking. Better abstractions can enable us to understand and reason about complex systems as the composition of simpler components, with clear interfaces. They can also lead to efficient production of high-quality network software, through code generation and code reuse. Network abstractions can take many forms, such as the precise statement of requirements, the design and specification of good interfaces between protocol layers, accurate models of how networks behave, or even qualitative principles underlying network architecture.
In this DIMACS working group, we plan to bring together "thought leaders" who are identifying new network abstractions, to have a working meeting with ample time for discussion. Our goal is to work toward a climate in in which researchers can make progress, by answering the following three questions:
To encourage active discussion, we plan to limit attendance to around 25 participants. Assigned speakers will survey the needs, issues, and prior work relevant to specific areas of networking.