Looking at the evolution of the Internet, in recent years we can identify a clear trend towards virtual communities. These communities are either 3D worlds like “SecondLife” and “World of Warcraft” or Web portals which enable the user to participate actively and richly. The HyperVerse project intends to leverage this trend in order to create a next generation Internet in which content is not consumed by navigating through Web pages, but by moving an avatar through a consistent 3D representation of the world. Interaction with content like objects, shops, auction platforms or other avatars is an integral part of this concept. Supported by advances in the field of human interface devices, interaction between users is likely to become more immediate and responsive, which will lead to increasing latency requirements. Experiencing such a next generation Internet is more immersive, intuitive and conforms naturally to the human mind’s cognitive capabilities.
The realization of the above mentioned vision implies several challenges. Massive and latency-sensitive real-time interaction requires a dynamic, scalable and self-organizing infrastructure. Traditional centralized Client/Server approaches cannot fulfill these requirements if applied on a global-scale while existing structured P2P overlays do not provide the facilities needed for the highly interactive scenarios we address. We propose a self-organizing infrastructure consisting of a massive number of federated reliable servers contributed by companies and organizations. Even individuals ought to contribute servers to the infrastructure since the average user bandwidth is likely to become symmetric and increases much faster than backbone and server bandwidth. Appropriate business models offering incentives for contributing such resources need to be invented. Furthermore novel concepts similar to the Torrent-technology ought to be applied in order to actively use the client’s network connection and unburden servers.
Existing approaches that focus on maximum interactivity suffer from severe scalability problems because they assume thin clients that do not contain any kind of object data. On the contrary, today’s massive multiplayer games are much more scalable because they distribute this kind of data at deployment time. Thus, only meta-information like object position, interaction between users etc. need to be exchanged at runtime. A downside of this approach is that it limits users with respect to creating and interacting with objects, since all world objects are predefined. In order to combine scalability and interactivity of both approaches we suggest a hybrid scheme using thin clients that become richer over time by using an aggressive caching scheme. With time, only relevant meta-data need to be exchanged for a working set of information. Unlike in today’s Internet, the above mentioned high interactivity requires true push-style communication since an object in the user’s field of vision can be modified anytime by anyone. Consequently, in the context of the HyperVerse project, an important aim is to investigate how push style communication can be supported by a future Internet infrastructure in a scalable way. Considering e.g. e-mail protocols or XML based feeds, even today’s Internet would benefit from such push style information dissemination.
HyperVerse is a joint research project at the Universities of Luxembourg and Trier, with focus on evaluating requirements for a 3D Internet as described above. This includes the definition of adequate interfaces and the development of a middleware for a massively federated server infrastructure based on widespread and established Internet standards like HTTP and Web Services. Based on a prototype 3D Internet browser and this server infrastructure, we intend to evaluate the concepts described above.