Imagine a system that is designed around the ability to adapt its appearance under the changing social needs and concerns of its actors!
In other words, imagine a system that is designed to work in ways similar to how societies themselves work.
While this may sound futuristic, even utopian, we do believe that it deserves a closer look. Social structures inherently express notions of who, what and why with respect to actions and tasks associated with the people and organizations that are immersed within these structures. Task- or goal-oriented computing has been chartered for a while to program solutions under certain given constraints. However, we assert that the underlying structures of networks and systems only inadequately reflect these notions of who, what and why in their present design and deployment. We further assert that pushing the notions of concerns and needs deep into the system will enable us to adapt the system under the changes that might occur in the future.
These changes are executed in runtime (i.e., upon need) rather than at design time (i.e., upon perception of the need through the designer) of the system. Moving the ability to adapt from design to runtime would introduce the required flexibility since it would minimize the dependency on the foresight of the system designer and maximize the ability of a given system to keep working under changed constraints.
We believe that such system-level ability of adaptation will eventually be enabled through embedding a plethora of intelligence, implemented in agents acting not only on behalf of the end user but also interacting with him. While some of this intelligence will need to be deeply embedded into the system, the underlying system also needs to provide the necessary hooks for such intelligence to be able to perform the required adaptation. With this, we envision that the resulting system and network design will accommodate virtually any social structure (and therefore business structure as well) without breaking the basics of the design. It will allow for executing these changes in runtime, within a single architecture, rather than requiring the design of a multiplicity of alternatives.
The aspirations outlined above go far beyond the scope of a single project like PURSUIT. It includes technical as well as socio-economic aspects around policies, HCI, information representation & reasoning, mediation but in particular also aspects in economics and regulation.
Within PURSUIT, we contribute to this larger vision by focussing on changing the routing and forwarding fabric of the global internetwork so as to operate entirely based on the notion of information (associated with a notion of labels to operate the fabric on) and its surrounding concerns, explicitly defining the scope of the information. While we do not embed the higher level semantics of information into the network, we intend to devise means that will enable the higher levels to embed concerns and social structures surrounding this information deeply within the architecture. This will be reflected by selected work items in our project such as forwarding and rendezvous.
As a background, the current model of IP used in the Internet establishes a routing fabric centred around a topological notion of the network, in which packets are delivered end-to-end between two explicitly named endpoints. Changes in this topology as well as changes in forwarding behaviour are difficult, albeit not impossible (e.g., firewalls). This leads to inflexibility not only on technical (see the photo example above) but also on business (e.g., inter-provider peering) level. Hence, the ability to adapt to changing needs on the technical, social and business levels is limited and usually achieved with a more evolutionary extension model rather than with a fundamental conceptual model for an adaptive network.
The envisioned operation on information, as formulated above, indicates an important move away from this endpoint-centric (data) networking model. While a model such as in IP networking enables a stream of data between two explicitly addressed endpoints (with total transparency as to the information represented in this sent data and the communication surrounding this exchange of data), the model envisioned in our aspiration elevates information onto the level of a first class citizen in the sense that data pieces are explicitly addressed and therefore directly embedded into the network, unlike in today’s IP-like networks. This explicit addressing of data instead of identifying endpoints consuming this data can be seen as the most significant step on this level of the overall architecture of moving from a data to an information view, i.e., establishing correlations between data through explicit (data) addressing. This is aligned with, e.g., Van Jacobson’s view on the evolution of communication from connecting wires (public telephone system) over connecting endpoints (current Internet) to connecting information (our aspiration above).