We believe that a broad class of future applications will span both the Internet and the telephone network because such multiplanar applications have several economic and architectural advantages over conventional ones. We also envision the close interlinking of the telephone network and the Internet to form a multimodal network. In this paper, we describe these applications and networks, outline their architecture, and present our experiences in constructing a prototype multiplanar application.

Computer applications, over the last fifty years, have evolved from programs running on a single mainframe, accessible only to users physically present at the mainframe installation, to distributed, multi-component systems whose parts communicate over a network. We can represent this progress with the following geometric metaphor. Programs running on a single machine are accessible only at single zero-dimensional ‘points’. For example, a program calculating ballistic trajectories on a mainframe receives its input and produces output at a single location. If we now introduce an underlying network, we can imagine ‘linear’ one-dimensional applications that require communication between two ‘points’ or hosts. This encompasses the client-server paradigm, where a single client and a single server cooperate to provide some service. An example of a linear application is a file server, where a stub program on a client communicates with a server to provide file access.

The next evolutionary stage is a ‘planar’ application, where the application requires simultaneous communication between multiple hosts. Two examples illustrate such applications. First, consider a cluster-based computation, where an application partitions computation among a cluster of machines. Distributed ray-tracing is a good example of this type of planar application. A second sort of planar application is a component-based application, such as a distributed database, where the application’s functionality is distributed among a set of communicating components.

Continuing this progression, the next obvious step is a three-dimensional application. What would the third dimension imply? Recall that the ‘plane’ in a planar application is formed by a communication network. Thus, in a three-dimensional application, one could imagine that each host would simultaneously be present on a several networks, allowing components to communicate using any of these networks. While such applications may become widely available in the future, the focus in this paper is a more restricted set of multiplanar applications: applications that span multiple components, and the different networking planes are linked by gateways. Thus, in a multiplanar application, every host can communicate with every other host present in any of the constituent networks, but only with the help of an intervening gateway.

We claim that such multiplanar applications are interesting not just in terms of their position in this evolutionary continuum, but also because they have significant economic and technical advantages over existing point, linear, and planar applications.

Along with the evolution of applications described above, there has been a parallel evolution in internetworking that we describe next. The world has three major internetworks: the telephone network, the Internet, and television and radio broadcast networks. Each network is independently being extended in breadth and depth. By an increase in breadth, we mean that each network is covering greater and greater geographical areas, and within each area, is achieving greater endpoint density. Cellular telephones, cable TV, and modem access to Internet are some of the technology drivers behind this trend. The three networks are also increasing in depth, that is, in the number of services available to an endpoint. Pay-per-view service in the television network, fax service in the telephone network, and Web-based services in the Internet are examples of increases in network depth.

While the telephone network and the television network have their advantages, such as guaranteed quality of service, and in the case of broadcast networks, a very low cost per endpoint, they are both much less flexible than the Internet in the creation of new services. Moreover, as the intrinsic bandwidth available in the Internet increases year after year, it is more and more likely to become the dominant network in the near future. This prompts the evolutionary trend of replacement, that is, a trend towards using the Internet to replace capabilities formerly provided by one of the other networks. For example, much of the revenue acquired by telephone companies from 800-number service is being replaced by catalog ordering over the Internet. Similarly, with real-time audio and video streaming, some of the functionality of the radio and TV networks is already being provided by the Internet.

Extrapolating from the above trends, one may come to the conclusion that eventually, the entire functionality of the telephone, television, and radio networks will be subsumed by the Internet. While this scenario has its merits, we believe that the existing installation base for these networks is so large, and the revenue that they already generate so substantial, that, at least in the near term, we will see the rapid deployment of multimode networks. These are networks built from multiple networking technologies, with interworking between disparate components provided by gateways. Such multimode networks allow users on one network to access resources on another network, thus leveraging the existing subscriber base. Clearly, such multimode networks are the substrate underlying the multiplanar applications described earlier.