Forty years ago—on December 5, 1969—the U.S. Department of Defense's Advanced Research Projects Agency (ARPA) connected four computer network nodes at the University of California, Los Angeles, (U.C.L.A.), the Stanford Research Institute in Menlo Park, Calif., U.C. Santa Barbara, and the University of Utah for the first time. This doubled the size of the embryonic ARPANET, the network that would grow over the years into the global nexus of interconnected computers we know today as the Internet.
Vint Cerf has been there from the beginning, from with his work co-developing TCP/IP (the communications protocols that the Internet uses to route information across different networks and hubs) to his present position as Google's chief Internet evangelist.
We sat down with Cerf, who is often called "the father of the Internet," to talk about why the ARPANET was built and how it grew to become the Internet, not to mention the pros and cons of social networks.
[An edited transcript of the interview follows.]
While we're honoring the 40th anniversary of the first time the four-node ARPANET was connected, what in your opinion was the single most important "event" in the development of the Internet?
Choosing a single most important development is incredibly hard to do because a lot of different things had to happen before the Internet could be deployed in the fashion it is today. ARPANET validated packet switching, which was important, because without that we wouldn't have gone down the path of toward the Internet. The idea was that you could grow a system like the Internet one network at a time and then interconnect them. In some sense the most important thing was the invention of the architecture protocols that enabled the Internet. Then came the implementation [on November 22, 1977] that interconnected different packet networks together. This included a radio network and a satellite network connected as well as the ARPANET. The most important thing in this three-network test was to show the TCP/IP protocol would link all three together.
What were your thoughts at the time, as a U.C.L.A. graduate student, when you realized ARPANET was going to be a successful endeavor? In which direction did you think the technology would go?
We could see how useful this was going to be almost immediately. The absolute first asset was remote access to people's computers. Remote access to someone else's time-share computer (most computers at the time were connected to mainframe servers and access was shared) was a very powerful tool. For the first time, you could use someone else's applications rather than just the ones you wrote yourself.
The impetus for this came out of the Advanced Research Projects Agency (ARPA), which was looking to make computer technology more relevant to the military. In the late 1960s every computer science department sponsored by ARPA said they needed the most advanced computers to do this work. So ARPA came up with the idea to connect computers for resource sharing. We immediately recognized the utility of that.
Although ARPANET was initially designed for the military (and only later on became available to defense labs, university labs and the general public), was it designed with an eye toward potential commercial applications, as well?
ARPA was responsible for developing systems for military command and control, but ARPA was also interested in pushing boundaries of computer usage for military applications. ARPA wanted to solve its immediate research problems, which led to remote access. By 1985, the National Science Foundation realized the utility of this for all research institutions supported by the NSF around the United States and built the NSFNET backbone for those institutions. (In 1988 the U.S. Federal Networking Council approved the interconnection of the NSFNET with MCI's mail system, a move that kicked off the commercialization of the Internet.)
Can you describe the changes you made to your networking protocols following ARPANET and why they were significant?
The purpose was to create a nonproprietary universal set of protocols. (Cerf and TCP/IP co-inventor Robert Kahn) said this would not be patented, it would not be constrained, it would not be controlled. We wanted to eliminate any excuse for not standardizing on our protocols globally. We didn't change anything about any of the existing networks. TCP/IP overlaid those networks. We relied on a device called a gateway to encapsulate our protocols in any protocols used by any new networks. The new networks had to have gateways (which are now called routers) that could encapsulate data so that it could be sent to other networks. The first commercial routers came out about 1986 and services came in 1987.
It's generally been said that the Internet wasn't designed to do the things we're asking of it now—that is, e-commerce involving sensitive financial information, the management of confidential documents such as electronic medical records, Voice over IP (VoIP) phone calls, etcetera—and that's one of the primary reasons that security has been such a challenge. What is your assessment of the security situation on the Internet?
It is true that when we started we didn't even know if this idea was going to work at all. In 1973 the only cryptographic technology we could get our hands on was classified. I was already working with the NSA [National Security Agency] to implement a secured version of the Internet that was classified, so I couldn't tell my friends about it. But security doesn't just come from cryptography, it also comes from making browsers that are less naive and don't download malware at the drop of a packet. The problem is that early on no one wanted to pay much attention to security. The business world just didn't care. They didn't see the threat, they didn't see the need, they didn't see any dependence on the Internet. It's only in the past decade or so that people have come to see this dependence.
I noticed that you have a Facebook page. What made you a fan of that social networking site, and which others do you use?
To be honest, I joined Facebook as an experiment. I accepted all invitations just to see how many people would ask to be "friends"—it quickly overwhelmed my time to process even the invitations and requests, let alone to actually go there and do anything. The various "applications" ("You've been bitten by an alligator.") and so on are just e-mail dressed up in awkward clothing. The user interface is rather clumsy in my opinion. I am annoyed by people that send messages via FaceBook because I get an e-mail telling me there is a message on FaceBook—so I end up processing two messages for every one sent.
I've seen modest utility in this system, but I have also heard that people do discover old friends this way. I think exploring the Internet's—and the Web's—ability to facilitate personal linkages is remarkable and expect to see additional social networking applications and services emerge. We have already discovered how quickly we become dependent on the Internet and it's applications for business, government and research, so it is not surprising that we are finding that we can apply this technology to enable or facilitate our social interactions, as well. Shared calendars are a tool for this as are Doodle polls—a very nice tool for finding out when multiple parties might be able to meet. So, while I am not an active Facebook subscriber and have not given in to Twitter, I appreciate that some of these tools have proven useful for a significant number of "Internauts".