New Internets
Today I met with Dr Guru Parulkar with the National Science Foundation on the Global Environment for Network Innovations (GENI). The project has a staggeringly difficult task: to reinvent the Internet bottom up. The theory behind this is actually quite simple: the current Internet architecture has powered more than thirty years of innovation, commerce and communication, but it is now in need of a review and there are obvious difficulties with security, address space limitations and basic networking functions. New innovations, like sensor networks and mobile clients, require a different structure and a new form of networking architecture.
Why a new Internet?
There is a great risk that the current Internet infrastructure actually creates a negative selection pressure on innovations in that it limits what can be done, for example in the sensor network space. To avoid this Dr Parulkar and the National Science Foundation founded GENI. The idea being to connect existing research on what the next generation Internet would look like and to develop a research facility for experiments that require a different, more versatile network design.
Sensor networking - where sensors with limited computing and communication power may be widely deployed at a cheap cost - will revolutionize much of what we know about science, communication and the net today. These many little sensors will be embedded, Dr Parulkar believes, in everything that costs more than ten dollars - starting a cambrian explosion of connectivity where the species of connected devices will explode in to multiplicity and result in a much more diverse network than today. This also requires new thinking on the part of the network designers, he notes.
Sensor networks are also interesting because they bring senses to the web. Hitherto the net has been numb, deaf, blind, without sense of smell or touch. With sensor networks - and some are already being tried out in military applications - all this will change. We will have a sensing network, a special kind of entity that will be able to monitor everything from pollutants in the air to the kind of perfume you are using (now, let's not get started on the privacy issues in a sensor network with DNA-sampling ability...).
The GENI facility - Dr Parulkar is careful to point out that it is not a testbed since a testbed is dedicated to one single idea - will be started in the year 2008 and be ready in five years. In a way it resembles a particle accelerator, but for network research: it is general facility with certain limitations that can be used by researchers to try out designed experiments in a controlled environment. The operating time of the testbed is presumed to be ten years.
Societal impact studies
When Dr Parulkar describes the project I ask him if they have decided to take into account societal impact issues as well, and he says that they plan to do so. It is, he explains, an integral part of the project. I admit that this really does not surprise me, but I still want to know why, so I asked him directly: why would you care about the societal impact? The guys who invented the Internet never did - they invented a then-superior form of networking system, that was all. They had no intention, capability or reason to discuss societal impact (and had they done so they would probably not have expected what actually happened!).
Dr Parulkar admits that this is the case, and when I say that there seems to be a political correctness to all the talk about societal impact in innovation systems he does not directly disagree, but he poses a counter-question. If we have reasonable grounds to believe that a technology can have a severe impact, should we not then examine this? And again, he notes, if a technology has a clear value proposition it will be adopted anyway, no matter what the societal impact may be.
Well, perhaps. The thing is that the discussions on societal impact almost always center on negative consequences, and the press these discussions get creates a negative incentive for investors and scientists to engage in different areas. And in order to discern the positive effects or value propositions the technology may need time to develop in a space where it is not hampered and our creativity is not limited by discussions about possible downsides.
The talk about societal impact may again create a negative selection pressure on innovation systems, forcing out innovations that could have great impact, but seem associated with possible (not probable, nota bene) drawbacks. The most obvious example is perhaps stem-cell research.
End of end to end architectures?
The next thing we discussed was the current focus on end-to-end network architectures. The old telephone network was a highly centralised and "smart" system in that all the intelligence actually was embedded in the network. The phones were dumd devices at the edges and nothing could be done with them to innovate. Anybody who wanted to innovate communication in the telephone networks had to work with the bottleneck of the central telephone company. When the Internet came, it reversed this almost completely. The Internet was in many senses an extremely dumd network and pushed intelligence to the edges, thus enabling one of the perhaps strongest and most diversified waves of innovations humanity has seen.
Now, will this be preserved in the next generation networks? Dr Parulkar thinks that the question is based on a faulty premise. It is not the case, he notes, that there has to be an either-or situation. With virtual overlay networking, for example, he explains, each end user at the edges can create their own network which should spur even more innovation. True - these may be intelligent or dumb networks, but that does not matter. What matters is that the network actually allows for yet another layer of innovation in restructuring the network.
The assumption behind the dichotomy between end-to-end networks and centralised networks is that we have to choose. But what if the user can choose to build virtual networks on the top of the existing infrastructure? Then he or she can best adapt their business models, research projects or other networks to the task at hand. The distinction between smart/dumb becomes a matter of choice made at the edges.
In summary
The fascinating thing about GENI is perhaps that it is all the Internet was not. It is a planned attempt to change the infrastructure of communication in the world. Perhaps it will even result, Dr Parulkar says, in a parallel network that overtakes the Internet much as the Internet supplanted the telephone network and it in turn supplanted the telegraph network.
We will see.