During the past 40+ years, numerous architectures were developed for network communication, including the ISO OSI reference model and its related protocol specifications and – of course – the Internet architecture. These network architectures all have been designed with some implicit assumptions about specific target applications and deployment scenarios. Among the most important assumptions are specific characteristics of the underlying network (= link layer) technologies, such as relatively short transmission delays, low error probability and the existence of end-to-end paths.

In certain advanced scenarios, these assumptions no longer hold. Examples of such advanced scenarios include networks with frequent connectivity disruptions, extremely long transmission delays or loose connectivity. Consequently, the existing network architectures fail to support communication in these scenarios, resulting in either significant inefficiencies or complete loss of connectivity.

Several localized, ad hoc solutions attempt to improve specific aspects of existing network architectures to better support these advanced scenarios, such as long/fat pipe extensions to individual transport protocols. These fixes can be successful in limited scenarios, but often lack broad applicability, i.e., they often address the symptoms of the issues instead of considering the causes. General architectural considerations are needed to approach the issues from a more fundamental and long-term perspective, rather than adding to a growing collection of short-term “patches.”

Disruption Tolerant Networking is a new area of research in the field of networking that deals with extending existing protocols or inventing new ones in a coordinated, architecturally clean fashion, to improve network communication when connectivity is periodic, intermittent and/or prone to disruptions.

Among the challenges of this field of research are large transmission delays. These may result either from physical link properties or extended periods of network partitioning. A second challenge is efficient routing in the presence of frequently disconnected, pre-scheduled, or opportunistic link availability. A third challenge is high link-error rates that make end-to-end reliability difficult. Finally, heterogeneous underlying network technologies (including non-IP-based internetworks) and application structure and security mechanisms capable of limiting network access prior to data transit are required in environments with very large round-trip-times. In some cases, an end-to-end path may not even exist at any single point in time. From a mobility perspective, DTN relaxes the “always on” paradigm, which would be extremely costly or even impossible to realize in challenged environments.

These challenges can decrease the reliability and performance of communications at essentially all layers of the protocol stack, ranging from packet-based forwarding and routing, to reliability and other features provided at the transport layer, to the application protocols (and applications) themselves. In addition, traditional mobility approaches may have to be revisited to accommodate users in networking environments prone to connectivity disruptions.

Numerous research activities over the past three years have focused on various facets of communications in challenged environments. Architectural concepts have been devised, prototype implementations were developed and research results are available from analysis, simulations and real-world experiments. The Dagstuhl seminar brought together researchers working in otherwise at least partly disjoint areas and established an intense dialogue across the variety of application domains.

In summary, this Dagstuhl seminar has sharpened the understanding of the very different perspectives from which researchers approach the problem space of disruption- tolerant networking, their assumptions and requirements, and the short- and longterm solutions they envision. This has broadened the view on DTN at large and contributes further issues to the present DTN research topics such as naming, security, service differentiation and efficiency. Assuming the traditional well-connected Internet architecture and its (interactive) applications as one extreme and the DTNRG architecture for purely asynchronous communications as another, the middle ground of mobile and partly (dis)connected operation may be approached from either edge. Future research will need to determine how far the DTNRG architecture can and should reach towards traditional Internet applications while maintaining its architectural integrity.