Net gains

20.03.2006
When Robert E. Kahn and Vinton Cerf created the TCP/IP protocol, they laid the foundation for today's Internet. Guru Parulkar, a 20-year veteran in the networking industry, is program director at the National Science Foundation 's computer and network systems division, where he is working on the GENI Initiative, a research testbed for the future Internet. Kahn and Cerf are also playing a role in the Internet's future: Kahn is chairman, CEO and president of the Corporation for National Research Initiatives, and Cerf is chief Internet evangelist at Google Inc. Computerworld's Robert L. Mitchell asked each of them to comment on the Internet's past, present and future.

Is the Internet today better than it was five years ago?

CERF: The Internet is larger than it was five years ago by a factor of at least five. It continues to function reliably, and the underlying systems have higher absolute capacity. It continues to expand to support new applications, including real-time interactive games, voice, collaborative tools and peer-to-peer applications.

New content is surging into the network, and search tools have become much more sophisticated. The network continues to support new applications that seem to be invented daily. Wireless access has proliferated, and mobile phones are increasingly Internet- enabled. Geolocation services are becoming more visible.

In the Internet, intelligence resides at the edge. In today's telecommunications networks, it resides in the network. As the two worlds converge, can these models continue to coexist?

PARULKAR: In the original Internet model, the network is [not] just a packet transport mechanism. There is a lot of intelligence inside the network, in the form of servers and devices. In the future, we have to think about balance. How much intelligence? Clearly, there is this recognition that we have to revisit this architectural principle.

We are talking about three scenarios. One is the Internet may borrow some of the ideas [of the telecommunications networks]. Another possibility is that a parallel infrastructure emerges. A third possibility is the telephone network becomes one and the same thing as the Internet. The telephone network is now merging with the Internet. The Internet may merge into the telephone system 20 years down the road.

KAHN: I tend to think of the Internet as not having edges. Increasingly, the network providers will see ways of doing intelligent things within their nets. The real question will be, to what extent can they work with application providers to do a better job of meeting user needs?

What has surprised you most about the way in which the Internet has evolved?

KAHN: The introduction of the personal computer suddenly meant that the number of machines that could be connected was thousands, tens of thousands, more. Some argue that we have close to a billion users on the Internet. That was a big surprise. I take it as a matter of pride that when we did the original design, we had ideas of scalability and modularity in mind. The fact that it did work was not a surprise, although we hadn't expected it.

The misuse of the Net was somewhat unexpected. I did not think people would misuse it as much as they did.

What are the challenges with using the Internet today for business communications?

CERF: The Internet continues to be challenged by viruses, worms and Trojan horses, [and] denial-of-service attacks against hosts and infrastructure. It has adapted to provide virtual private network access for businesses, but firewalls and network address translation devices sometimes interfere with VPNs, packet voice services and the like. Authentication of users is still a challenge, and there is a need for more end-to-end confidentiality. Deployment of IPv6 is still slow in most of the network. The problems are severe and lie at several layers in the protocol hierarchy as well as in the quality of the operating system software that is so readily penetrated. Operating system and application system designers have not taken into account adequate measures for authentication for purposes of access control, leaving many systems open to exploitation.

KAHN: Security was something that we thought about in the early days, but there were so many obstacles to trying to deploy strong security that we just didn't go down that path. I wouldn't say there's no security but ... it could be better. I would like to see the whole issue of identity management addressed so you can authenticate information when you get it. But I don't know [that] that's a limitation. It's just an aspiration for something that could be better.

PARULKAR: If you look at mobile wireless, the Internet was not designed for this degree of mobility. The IP address is identical to an attachment point where you interface to the network. When people are constantly moving, your attachment point is constantly changing. So there is this issue of, can you continue to use IP addresses with this architecture, or can you do better by decoupling those two things? You may have to have another level of addressing.

How will those problems be fixed?

PARULKAR: Security problems, spam, robustness -- these are not problems you can fix the way we are doing it, by adding more and more boxes, like firewalls and IDSs. Some are inherent to the architecture. The assumption of the Internet was that all of the traffic was friendly. We know that doesn't work. There are no easy architectural mechanisms to fix it.

CERF: I think some basic rethinking of security and some built-in primitives are needed in the architecture to tackle this problem adequately. Adding features won't, in themselves, solve the problem without some serious reworking of the design of the system.

What do you think the Internet will look like in 10 years?

CERF: Two [billion] to 3 billion users, more devices on the network than people, huge archives of entertainment content, lots of third-party services to manage Internet-enabled devices, lots of wireless access, lots of high-speed fiber/cable for consumers, mobiles fully Internet-enabled, much more refined search services with significant vertical components, much more collaborative interaction, significant financial transactions via the Net, more diverse online advertising, and an operational two-planet interplanetary Internet -- Earth, Mars -- with plans to extend to the outer planets.

KAHN: It's a medium of social interaction. Of business interaction. Of information access. It's also very fragile in that it requires cooperation around the world for it to work. I hope that will continue.

The potential for business-to-business interaction [hasn't] been exploited to nearly the extent that it could be. Today, most people access business Web sites to see what companies are offering, maybe place orders. But that's just the very early start.

The ability of the Net itself to facilitate bringing together virtual organizations will be very significant. It isn't quite here yet, but I think that's one of the next steps we'll see.

PARULKAR: The research community has to look beyond the Internet, and that is what GENI is. GENI has two parts. One is research. The other is an experimental facility in which [researchers] can demonstrate and deploy technologies at scale. There are a half-dozen experimental infrastructures that the NSF has supported. PlanetLab is a kind of a global experimental infrastructure. It supports what are called virtual overlay networks on top of the Internet.

The paradigm is one physical infrastructure that could be used by my people's teams at the same time to deploy their own ideas -- a physical infrastructure that you can slice into virtual slices. Within that slice, they can deploy their own protocol stack and services and demonstrate that. The power of it is you can give different people a slice of a resource, and they think the whole resource belongs to them. Everybody agrees it's a great means for an interesting research framework. But some people are saying maybe that's a great operational framework as well.

What users want is to be able to create their own virtual network with its own behavior, security, robustness, QoS. Rather than point-to-point transport, if you give them the mechanism by which they get their own virtual network, corporations will be happier. Providers are able to sell something which is more value-add than VPNs. They can provide even more sophisticated services to customers.

Sidebar

Timeline: Historical nodes on the Net

1957: The U.S.S.R. launches Sputnik, the first man-made satellite. In response, President Eisenhower creates the Advanced Research Projects Agency (ARPA) within the U.S. Department of Defense.

1962-68: Packet-switching networks are developed.

1969: The Defense Department commissions the Arpanet for research into networking, with the first node at the University of California, Los Angeles, followed by nodes at Stanford University; the University of California, Santa Barbara; and the University of Utah.

1971: The first e-mail is sent.

1973: The first global connections to the Arpanet are established, with nodes at University College in London and Royal Radar Establishment in Norway.

1974: Cerf and Kahn publish "A Protocol for Packet Network Intercommunication," which is now recognized as a groundbreaking document describing how to connect several different packet-switching networks.

1976-82: Kahn and Cerf lead the way in the development of the TCP and IP protocols for Arpanet; the protocols will become the backbone of the Internet.

1984: The number of hosts exceeds 1,000. The Domain Name System is introduced.

1989: The number of hosts exceeds 100,000.

1990: Arpanet ceases to exist, symbolizing the commercialization of the Internet. The number of hosts exceeds 300,000.

1991: The World Wide Web, developed by Tim Berners-Lee, is released by CERN. The Web provides a distributed hypermedia system for the Internet.

1993: The graphical browser Mosaic is developed by Marc Andreessen and his team at the National Center for Supercomputing Applications. The White House and the United Nations come online.

1994: Local communities, the first being Lexington and Cambridge, Mass., come online as the Internet begins to touch society in many ways and at many levels.

1998: Microsoft jumps into the network service provider and browser market.

2006: The Internet is ubiquitous, with wireless access and millions of nodes, and is becoming a primary transaction vehicle of the global economy.