Design Space, Nothing But Net
Interview with Vinton Cerf
Dr. Vinton Cerf is often cited as one of the internet's founding fathers.
|Dr. Vinton Cerf, among founding fathers of the internet, contributor to Interplanetary Internet study group. "Little did we know thirty years ago that this research effort would spawn countless initiatives.." Credit: revistapoder.com
Cerf co-designed the transfer protocols (TCP/IP) while working on the Defense Department's earliest implementation, called ARPANET. He serves as chairman of the board of the Internet Corporation for Assigned Names and Numbers (ICANN) and sits on the Board of Directors for the Endowment for Excellence in Education, Folger Shakespeare Library, Gallaudet University, the MarcoPolo Foundation, and numerous corporations. He is a distinguished Visiting Scientist at JPL. In December 1997, President Clinton presented the U.S. National Medal of Technology to Cerf and his partner, Robert E. Kahn, for founding and developing the Internet.
Dr. Cerf is also a frequent contributor to many internet societies, and occasionally publishes in the otherwise formal format, called Request for Comments (RFC). In one such 1994 comment, he imagined as a kind of reverse time capsule. Cerf began to describe yesterday, from the future looking back to today. That vision took shape as what has become a project of his--building the Interplanetary Internet.
In those fascinating 1994 April Fool's emails, Cerf first outlined the requirements for a future interplanetary internet, one that has the ability to email Mars, the moon, and your local terrestrial coffee shop. On April 1, 1994, he described how it might all work in hindsight, with a literary exchange of emails from the future. From a Martian Exobiology Lab in the year, 2023, the exchange illustrated how it might begin.
Appropriately the current blitz of Martian landers seems to have shared some complex communication channels of their own, since for the first time they involves the use of a satellite constellation on another planet. As noted by Dr. Ed Weiler, NASA Associate Administrator for Science, one first for this round of missions that has paid off early, was the world's only interplanetary satellite communication network. First images arrived via orbiters around Mars, as unique relay points for storing and transmitting the data-rich pictures only hours after touching down. The large amount of data -- nearly 100 megabits -- transmitted from the Spirit rover in a single relay session through NASA's Mars Odyssey spacecraft was, according to mission manager, Jennifer Trosper, "like getting an upgrade to our Internet connection."
Astrobiology Magazine's web architect, Dr. David Noever, had the chance to talk with Dr. Cerf about his vision of the future and what it might take to wire the solar system.
Astrobiology Magazine (AM): Before working on ARPANET in the late 60s and early 70s, you had summer jobs in the heyday of the early space race with North American Aviation (now Rockwell). Can you describe what those jobs entailed for you and your impression of how the challenges of 1960s information Systems at Rockwell are most different from today?
Vint Cerf (VC): For the most part I worked on data collection and analysis associated with engine testing (in the Santa Susana facility in Simi Valley, California) or with computations associated the design of the engines themselves. In the earliest period (around 1960) I used a machine calculator and filled out endless spreadsheets.
A couple of years later I was at the Space and Information Systems Division programming analytic routines to process the data associated with the Apollo program. The systems then were big centralized Iron (otherwise usually known as IBM).
Today one would find mainframes, clusters and even some grid-like computing environments to say nothing of very powerful personal computing engines. Computations by hand took a long time and programming involved many hours of turn-around time. A far cry from today's real-time computing environments that support computational exploration of possibilities in design space.
AM: In December 1985, you authored a very short Request for Comment (RFC 968) for the Network Working Group entitled 'Twas the Night Before Start-up'. The RFC discusses problems that arise and debugging techniques used in bringing a new network into operation, with some clever turns like:
'Twas the night before start-up and all through the net, not a packet was moving; no bit nor octet...
He spoke not a word, but went straight to his work, fixing a net that had gone plumb berserk;
And laying a finger on one suspect line, he entered a patch and the net came up fine!'
How did you come to pen this memorable description of jump-starting the early internet?
VC: Actually, these words were penned in contemplation of the start up of the MCI Mail system [the first commercial email service to be connected to the Internet] in the Fall of 1983. There was an important network component to the system. The events described did not actually happen but they are not unlike problems arising in anything dependent upon software.
|SETI@home uses the idle time of over three million personal computers to sift through radio data for signals from extraterrestrial civilizations. Carl Sagan's widow, Ann Druyan: "That is about as democratic an approach to science, this notion of massive distributed computing, as any. It's exactly on the theme that is so dear to us, which is the idea of making it possible for all of us to participate in the experience of science. We are in the process of becoming an intercommunicating organism".
Also anyone that has done any serious debugging knows you can stare at code for hours and not see it - then some SOB shows up and in 30 seconds points to the one thing that's wrong. ("and then I hit him, officer...").
AM: And the origin of the closing reference in the poem to the root cause of bringing down the TCP/IP network using a 'off-by-one index'?
VC: As to the "off by one" bug - this is a very common programming mistake when dealing with arrays - especially when the indexing starts with "0" (zero) - one's intuition about the "end" of the loop count can be fooled by the fact that the loop is N+1 steps if it starts at zero and ends at N.
I've made my share of these bloopers.
AM: In September 1989, on the 20th anniversary of the ARPANET, you wrote another spoof (RFC 1121) on Shakespeare's character-- Rosencrantz --in Hamlet, called 'Rosencrantz and Ethernet':
"All the world's a net! And all the data in it merely packets come to store-and-forward in the queues a while and then are heard no more. 'Tis a network waiting to be switched!
To switch or not to switch? That is the question. Whether 'tis wiser in the net to suffer the store and forward of stochastic networks or to raise up circuits against a sea of packets and, by dedication, serve them."
If the world is now a net and the question is 'to switch or not to switch', what is the answer today, as we approach the 35th anniversary of ARPANET?
VC: The answer has blazed its way across our terrestrial universe. Conventional circuit switching applications are migrating to packet mode of operation - not necessarily all of them and certainly not all at once. But the strong showing of Voice over IP (VOIP) is an indicator.
2003 marked a very visible trend in the traditional telecom markets to migrate significant quantities of conventional circuit-switched applications to packet (or "datagram" mode). This has been an age old debate between the proponents of the two kinds of switching and the packet guys seem to be coming out on top. However, for really heavy continuous and largely invariant flows, circuit switching is quite efficient.
AM: On April Fool's Day in 1994, you penned a letter rescued from a time capsule and sent from the year 2023, entitled "A View from the 21st Century" (RFC 1607). The letter is from an email address 'email@example.com' with the subject line 'Hello from the Exobiology Lab!'. It begins 'I just wanted to let you know that I have settled in my new offices at the Exobiology Lab at the Interplanetary Space Exploration Agency's base here on Mars. The trip out was uneventful...'
|Layered martian terrain in painting by Bill Hartmann (left), orbital image from Mars Orbital Camera (right).
© William K. Hartmann
The ensuing three-way email conversation circles between France, the Moon and Mars, and is a joint research project documenting the history of the internet. This also mentions the Interplanetary Internet. Can you describe how this notion of space connectivity took hold in your imagination?
VC: I have to thank you for drawing this to my attention. I hadn't thought about the link between my late 1997 awakening to the possibility of an Interplanetary Internet and the earlier fanciful sketch.
The 1994 sketch was really intended to suggest to readers that the technical debates of the 1990s, intense as they were, would be resolved and the Internet would indeed expand not only on our home planet but throughout the solar system. I began speaking about the challenge of designing an Interplanetary Internet in late 1997.
|'For the first time in human history', said NASA Associate Administrator for Science, Ed Weiler,'we have an interplanetary communication channel over another planet. Data moved from Spirit [on the surface] to Odyssey [in orbit] to the Deep Space Network [on Earth, positioned in Canberra, Australia, Goldstone, California and Madrid, Spain].'
By good fortune, one of the engineers in my group, Greg Miller, had worked with a team at the Jet Propulsion Laboratory earlier in his career. He heard me making Interplanetary network noises and introduced me to Adrian Hooke, a leading space communication standards advocate, and when we met at MCI in the Spring of 1998, we were finishing each other's sentences.
Adrian had a small team already hard at work on extensions to TCP/IP to help it adapt to some of the serious impairments in the communications environment of deep space. Long delay is only one of the problems. Limited power, discontinuous channels (satellites disapper behind the planet...), asymmetric data and error rates, etc. all combine to make deep space communication a remarkable challenge.
It was soon apparent as we worked on the issues that the standard TCP/IP protocols or even their adapted versions simply would not perform satisfactorily in such a long haul environment. From Earth to Mars can take up to 20 minutes at the speed of light! Adrian and his colleagues and I worked from 1998 to the present on a new design for these tough requirements. The work is starting to bear fruit.
Some support was provided to us by the US Defense Advanced Research Projects Agency (DARPA) who had funded the original Internet design and where I worked from 1976-1982 as a program manager for the Internet project.
We are currently discussing missions planned late in this decade such as an orbiting communications satellite operating near Mars and also missions closer to home - on the Moon, in response to President G.W. Bush's recent proposals.
AM: What is the current status of the Interplanetary Internet?
VC: We have a suite of protocols designed and documented, and have released open source software to allow interested parties to test the ideas. We have discovered that the ideas can be adapted to some difficult problems here on Earth ranging from tactical military communication to sensor network systems.
Plainly these potentially mobile environments are also found, e.g., on the surface of Mars or other planets where we may send robotic observers and, someday, perhaps, people. While we have not yet flight tested the entire suite of protocols for the so-called InterPlaNetary Internet, we have some of the lower layers in use in current space missions.
|Mars Odyssey, named after Sir Arthur C. Clarke's 2001: Space Odyssey series is part of the constellation of orbiters currently relaying data on interplanetary connections.The large amount of data -- nearly 100 megabits -- transmitted from the Spirit rover in a single relay session through NASA's Mars Odyssey spacecraft was, according to mission manager, Jennifer Trosper, "like getting an upgrade to our Internet connection."
With good luck we may have the opportunity to test these ideas more fully aboard the 2009 Mars Telecommunications Orbiter. We are also interested in testig these ideas terrestrially, in military and/or civilian sensor networks.
AM: Who else is working with you on this project?
VC: The Interplanetary Internet team includes:
- Adrian Hooke - JPL;
- Scott Burleigh - JPL;
- Leigh Torgerson - JPL;
- Keith Scott - MITRE;
- Bob Durst - MITRE;
- Howard Weiss - SPARTA;
- Kevin Fall - INTEL Labs;
- Juan Alonso - Swedish Institute for Computer Science;
- Eric Travis - GST.
We have also drawn on stellar help (no pun intended) from UCLA, USC/ISI, MIT, Univ Delaware, Berkeley, among many others.
AM: The current round of Mars landers all share a common broadcast capability, S-band radio. How would an internet transmission likely travel as packets over such enormous distances?
VC: Yes, in fact they already often do - a protocol for packetizing the transmissions arises from the standards work of the Consultative Committee on Space Data Systems (CCSDS) forms the underlying basis of packet transport for the Mars rover program.
Shared common channels were a staple of the University of Hawaii's so-called ALOHANET. Packets can fill the transmission channel in the same way that cars can fill a road: one after the other. Despite the long round-trip times, the packet method of communication will allow many sources to be multiplexed on a common transport mechanism. As long as the protocols don't involve a lot of "back and forth" handshaking, the system can be efficient assuming the "pipe" is full of packets.
The successful Mars missions ignites a great interest in me in the creation of a solar system observation program to put many platforms in orbit and on the surface of moons and planets.
|First sunset on Mars, where each day, or sol, is slightly longer than an Earth-day
Credit: NASA/JPL/Cornell University
Such continuous data gathering would not only teach us a great deal about our own solar system but it would also develop information we will need to plan serious, long range and long term human exploration. Robotic exploration will be the key to longer term manned exploration: we must have a good idea what we have to deal with before we send humans on extended exploratory missions around the Solar System.
Dr. Cerf concluded his 1994 April Fool's emails from a base station on Mars--in the year 2023-- with the sign-off:
The letters end here, so we are left to speculate about many of the loose ends not tied up in this informal exchange. Obviously, our current struggles ultimately will be resolved and a very different, information-intensive world will evolve from the present. There are a great many policy, technical and economic questions that remain to be answered to guide our progress towards the environment described in part in these messages.
It will be an interesting two or three decades ahead!
Related Web Pages
A View From the Future
MIT Inventor of the Week
Long, Strange Trips
PBS: Is Science Fiction Science? Michael Crichton, David Brin, Octavia Butler
A Perfect World I: Tyson
A Perfect World II: Richardson
A Perfect World III: Goldin
A Perfect World IV: Venter
A Perfect World V: Hendricks
A Perfect World VI: Fuller