Vint Cerf is one of the privileged few computer scientists who can claim to have helped change the course of history. His pioneering work with Robert Kahn led to the development of the TCP/IP networking protocols, which would underpin the Internet. Both Kahn, who at the time was on the payroll of the Defense Department, and Cerf received government funding for their research. Ditto for much of the collaborative work undertaken by Cerf and the other now-legendary names involved in creating the myriad integuments of the Internet. (You can find a good narrative of their labors at this site managed by the Internet Society.)
However, the role played by the U.S. government in the history of the Internet has suddenly turned into a topic of debate. On Monday, former Wall Street Journal publisher Gordon Crovitz offered a decidedly revisionist take with a column titled, "Who Really Invented the Internet?" In the piece, Crovitz cast doubt on the assertion that the U.S. government deserved credit for helping create the Internet. He described the claim as an urban myth. Instead, Crovitz, who said that Xerox deserves the credit, refracted the question through a different lens: "It's important to understand the history of the Internet because it's too often wrongly cited to justify big government," he wrote.
In an e-mail Q&A, Cerf, nowadays working as an Internet evangelist at Google, offered his vivid recall of what it was like to be present at the creation and a different recollection about what it was like during those heady days.
Q: In his Wall Street Journal column, Gordon Crovitz writes that the federal government's involvement in the creation of the Internet was modest. Does that jibe with your recollection?
Vint Cerf: No. The United States government via ARPA started the project. (Bob Kahn initiated the Internetting project when he joined ARPA in late 1972. He had been principal architect of the ARPANET IMP (packet switch) while at BBN.
Bob invited me to work with him on open networking in the spring of 1973. We also both worked on the ARPANET project starting in 1968. ARPANET was funded through 1990 by ARPA and other USG agencies. The Internet work was funded from 1973 to about 1995 (and beyond) by ARPA, NSF, DOE, NASA among others. It took 10 years of work to get from the original paper published in May 1974 to the rollout of the Internet operationally on January 1, 1983. It combined the ARPANET, MILNET, some number of Ethernets, two Packet Radio networks, the Packet Satellite network, and other local networks in England and Norway. Note that University College London and the Norwegian Defense Research Establishment were involved in the implementation and testing of TCP/IP as was Stanford and BBN.
The NSF got very involved in 1985 and this led to the design and implementation and subsequent expansion of the NSFNET that became a major backbone for academic access to the Internet. NSF also sponsored more than a dozen intermediate level regional networks. By 1986, router companies such as Cisco and Proteon were selling to academia and the military and to USG-sponsored networking users. By 1989, three commercial Internet service providers were in operation: UUNET, PSINET, and CERFNET.
By 1992, the Boucher Bill make it permissible to carry commercial traffic on the U.S. government-sponsored backbones (notably NSFNET; ARPANET had been retired as of 1990). About that same time, Tim Berners-Lees' development of the World Wide Web protocols at CERN, 1989-1991, had gotten the attention of Marc Andreesen and Eric Bina at the National Center for Supercomputer Applications leading the development of the graphical MOSAIC browser that led Jim Clark (founder of Silicon Graphics) to start Netscape Communications with Marc.
NSF retired the NSFNET in 1995 but funded Network Access Points to assure commercial network and academic and government network connectivity and then went on to sponsor a new high speed research network called vBNSNET built and operated by MCI. That was joined by the Abilene network developed by the Internet2 academic consortium.
The U.S. government, including ARPA, NSF, DOE, NASA among others absolutely facilitated, underwrote, and pioneered the development of the Internet. The private sector engaged around 12 years into the program (about 1984-85) and was very much involved in powering the spread of the system. But none of this would have happened without this research support.
Any areas where the government fell short?
If the U.S. government were to be faulted, it would be because a different part of the government -- NIST as well as DoD's DISA (then DCA) -- abandoned the Internet in favor of the Open Systems Interconnection system that began in Europe in 1978 just as the Internet's TCP/IP was being finalized and standardized and put into use. For nearly 15 years (from 1978 to 1993) there was a pitched standards battle between TCP/IP and OSI and some parts of the USG (NIST and DCA/DISA) the USG took the OSI side after ARPA, DOE, NASA and DOE invested successfully in the development of the TCP/IP protocols. Ultimately, NIST was asked by the Internet Society in 1992 to evaluate TCP/IP vs OSI and after a year, NIST's Blue Ribbon Committee agreed that TCP and the OSI TP protocols were equally acceptable for use in USG applications leading to rapid spread of the Internet (which probably would have happened anyway given the modest degree to which OSI was ever implemented).
Crovitz also writes approvingly of this blogger's quote (from 1999): "The Internet, in fact, reaffirms the basic free market critique of large government. Here for 30 years the government had an immensely useful protocol for transferring
information, TCP/IP, but it languished...In less than a decade, private
concerns have taken that protocol and created one of the most important
technological revolutions of the millennia." Since you developed the TCP/IP
protocol, I'd like to know your reaction.
Cerf: I would happily fertilize my tomatoes with Crovitz' assertion.
It's hard, if not impossible, to prove a counter-factual. But let's assume
for a second that the government had not become involved through ARPAnet and
TCP/IP. What happens to the subsequent narrative? Does the Internet happen
or does it get delayed -- or was Crovitz right that there were enough "fired
up imaginations" so that "by the 1960s technologists were trying to connect
separate physical communications networks into one global network--a
You might have ended up with OSI. Many engineers considered this to be an overly complex design and it was not very much implemented.
Crovitz lays out a case where he says "full credit" for the Internet ought
to go to Xerox PARC Labs. Should it?
No. Xerox gets and deserves credit for great work: Ethernet (that was stimulated by the ARPA-sponsored radio Alohanet at University of Hawaii); the laser printer; the ALTO personal computer; Xerox Network System and PARC Universal Packet (PUP) -- this system was kept proprietary and while its developers occasionally gave hints as to its operation in my Stanford seminars where TCP/IP was being more elaborately developed, they did not go very far commercially. XNS ended up as IPX in the Novell systems and eventually died in favor of TCP/IP. XEROX did link homogenous Ethernets together but the internetworking method did not scale particularly well. NSFNET pushed the boundaries of scale leading to the development of the Border Gateway Protocol, a successor to the Exterior Gateway Protocol originally used in the Internet (not counting some experimental gateways even earlier). The Internet was designed to link heterogeneous network together from the start (ARPANET, Packet Radio Net and Packet Satellite Net).
Looking at the question more broadly, is there a way to assign
responsibility to one organization over another when it comes to something
like the creation of the Internet. Or because of the complexity, is that
really a fool's errand?
You can look at timelines to see the development. While "networking" was in the air in the late 1960s and early 1970s the specific Internet design came from the joint work of Bob Kahn and me. We had the benefit of the ARPANET experience, meeting with Louis Pouzin and his group at IRIA (Cyclades/Cigale networks), participating in the International Network Working Group (INWG) that I chaired initially, other network researchers in England, Norway, Italy, Germany, Japan, France, etc. After our initial paper was published, detailed design was conducted at Stanford during 1974 and implementation started in 1975 at Stanford, BBN and University College London. After that, a number of other institutions, notably MIT, SRI, ISI, UCLA, NDRE, engaged heavily in the work.
Until now, I wasn't aware that there was a debate over the origins of the
Internet. As a technologist, do you worry that the politicization of the
tech industry's history might impact the federal government support for
future computer research?
I hope not. Articles like Crovitz' distort history for political purposes and I hope people who want to know the real story will discount this kind of revisionist interpretation.