RDVP Seminar: Len Kleinrock, "Father of Modern Data Networking" (1/12/2005)

Len Kleinrock, computer science professor at UCLA and the inventor of packet switching and a key figure in the creation of the Internet, spoke at the seminar. He combined a history of the internet along with his views on the next stages of network development, and a few entertaining stories for good measure. He asserted that the explosion of the internet was a result of the desire of hundreds of millions of people to access and contribute information and communicate with each other combined with the culture that had been established in the early days and had persisted for much of its growth. The key elements of that culture include:

• open research
• open architecture
• shared ideas
• no overbearing, centralized control
• trust (though he felt that trustworthiness had eroded on the network today)
• group consensus

He talked a bit about the network protocol stack, judging the creating of the IP layer as the critical layer of abstraction providing the application layers above it the ability to rely on whatever network service layers happened to be below it. And it had focused simply on addressing and quality of service.
He offered a chronology of key events in the creation and growth of the internet. I’ve included only a small subset below, but the surprising thing to me was how recent it all was. Just realizing that the inventor of packet switching is still an active researcher puts some of it in perspective. A couple other “events” in the early-to-mid 90’s that I still remember clearly as “not that long ago” showed up as “historical” events that marked some turning point in the development/commercialization of the internet.

• April 1962: First paper on packet switching
  
• 1963: Kleinrock joins the UCLA faculty
  
• 1967: Wes Clark argues for network control being separated from host computers (routers)
  
• 1968: Bob Kahn at BBN starts building, selecting UCLA as the first node.
  
• July 3, 1969: Press release where Kleinrock anticipates the provision of network computing as a utility.
  
• 1969: UCLA connects from the host machine to the switch
  
• October 29, 1969: First message sent from UCLA to SRI “Lo” (actually the start of Login, before the SRI machine crashed)
  
• 1972: Roy Tomlinson introduces e-mail, the application that becomes the bulk of traffic
  
• 1983: Standardized on TCP/IP
  
• 1988: Robert Morris writes the first worm, spreading beyond his intent. (Robert Morris overlapped with me at Harvard, where he was a research associate. It was hard to believe that this was nearly 17 years ago, or almost half the life of the internet.)
  
• 1991: NSF, which has jurisdiction of the Internet, opens it up to commercial activity, sowing the seeds of yet more growth.
  
• 1994: Cantor & Siegel send the first spam, announcing the end of the green card lottery. (This was also a trip down memory lane. Sure, I remember that! I remember the firestorm of outrage that followed. Spam is really only 10 years old? It seems like such a common part of life now that it MUST be older than that. There are whole industries around spam filtering.)
  
• 1996: The number of email messages in the US surpasses the number of physical mail messages.
  

Kleinrock also talked about some of the bad assumptions that they made that are causing headaches now. The key one was that they assumed a person, his machine, his physical location, and his IP address were all equivalent: that is, office-based computing. As people travel with laptops, share computers, have multiple computers, etc, this simplification has caused problems.

This mobility and proliferation of machines has also set the stage for the next 3 phases of internet development, which he described as:

1. Nomadic computing
   
2. Embedded Technologies for Smart Spaces
   
3. Ubiquitous Computing
   

Assumptions about trust within the environment meant that now that trust has eroded, there’s no easy way to either perform (or add the capability) for strong authentication, protection against malicious attacks, provide strong end-to-end encryption.

He mentioned a current project to address some of the issues of nomadic computing, called Spheres of Influence at UCLA, which handles some policy management issues for networks and users. He also talked a bit about the failure modes of systems, where some (especially complex systems) give no hint that they’re not working correctly until they fail catastrophically. This is an especially big risk when we delegate important tasks and don’t have adequate measurement. (Kleinrock established and ran the Network Management Center at UCLA from the very first days of the internet.)

Along the way, he told some fun stories about

• His blackjack system
  
• Making timings of the roulette wheel at Vegas
  
• Samuel’s checker program programmed to lose and yet still playing a dominating game until the point where it sacrificed everything
  
• Trying to get distributed systems to learn a pay-off matrix that relies on the outputs of all the agents in the system, but with no communication between agents (the agent’s learning can only take into account the payoff that it receives from the system)
  
• Challenged to find an obscure landing craft from WWII on the internet, finds not only the history of it and a picture, but a picture that includes the person who issued the challenge!
  

Afterwards, Durga and I gave Len a ride to the San Jose airport. Durga took this shot of him in my car
.