Schooled By Vint Cerf

By Brian Boyko, Editor, Network Performance Daily.
Perhaps more than anyone else, Vint Cerf is the creator of the Internet. He co-designed the TCP/IP protocol when he was working at DARPA, and currently serves as the chairman of ICANN. Of E-Week’s Top 100 Most Influential People in IT, he ranks at #20. Among the geek set, he is the stuff of legend.
So I was honored to find out that he posted a reply to one of Network Performance Daily’s Tuesday links, where I made fun of the idea of an Interplanetary Internet [additional info at Wikipedia] – an idea that Mr. Cerf has been working on with NASA’s Jet Propulsion Laboratory.
I was also humbled because that reply was:

Re: Space Internet – You have completely misunderstood the design. We DON’T use TCP or even IP for the long haul. We use DTN – see RFC 4838 and also www.dtnrg.org for details. We use the TCP/IP protocols for onboard spacecraft and also in low delay environments on the surface of planets but DTN for the interplanetary components.

So, on one hand, Vint Cerf read what I wrote. On the other hand, he pointed out – correctly – that I dismissed the idea of Interplanetary Internet far too quickly.
Furthermore, one of the things that I wrote in that article was that it would take about a month and a half for a triple-handshake connection between Earth and Mars. This was, quite frankly, due to a dumb math error on my part. The actual time it takes, based on an Earth to Mars distance of 400 million km, and a speed of light of roughly 300,000 km/s, is roughly 1,333,333 milliseconds – or 22 minutes – 66 minutes for a triple handshake. High latency indeed, but not the two weeks per pass I had originally thought. (Never again will I make fun of the probe that crashed into Mars because of problems converting imperial measurements to metric.) The post has since been strike-through corrected.
I can only say that “a little knowledge is dangerous” and led me to make poor news judgment.
So I offer a personal apology, to Dr. Cerf, to the people at the NASA Jet Propulsion Laboratory, and most importantly, to the readers. Following Dr. Cerf’s links has led me to some information that I now find both very interesting and very relevant to people following networking and network performance.
RFC 4838 for Delay-Tolerant Networking Architecture and www.dtnrg.org are indeed very good sources of information about this protocol.
There are many benefits of adding interconnected networks to space technology. According to this Cisco press release, those benefits include improvements in interoperation and security with ground components, as well as additional flexibility in space design.
But the practical applications of the DTN protocol go far beyond space applications as the system is designed to be used among performance-challenged environments. Performance challenged environments include disaster recovery scenarios, the developing world, and various military applications.
The DTN protocol is designed to work with very large delays, including natural propagation delay, such as, for example, the delay between Earth and Mars. DTN is designed to work where standard Internet fails, by using variable-length messages instead of limited-sized packets, by using storage within the network to support store-and-forward operation over multiple paths and long timescales that do not require end-to-end reliability. This requires that space routers have the ability to retain data over a much longer period of time than a standard router, and to be able to retain that data even after a reboot. The DTN protocol also has added security (the last thing you want is Ernst Stavro Blofeld hacking into your spaceship-to-shore communications) and built in classes of service.
Applications that use DTN will also be designed to minimize the number of round-trip exchanges (and hopefully some of those development techniques will make their way into WAN environments.)
So the high-latency, low-reliability environment of outer space requires new protocols and new ways of thinking. And I can’t think of anyone else more qualified to develop that protocol than Vint Cerf. After all, he’s already got experience in the protocol development area.
Mr. Cerf also directed me to Adrian Hooke at NASA. I plan to ask him some questions about the technology later this week, and look forward to sharing that information with you in a future post.