By: Henri Wohlfarth
Date: May 7, 2006
Question: Lixia, you attended the first IETF Meeting. How did you get involved in the IETF?
Lixia: I was at MIT working with Dave Clark. He was the first IAB chair from 1981 – 1989. That were exactly the same 8 years I was at MIT as his graduate student. There used to be a few meetings every year with the whole network team in the same room. As the community was growing large, around 1984 or so Dave Clark and Berry Leiner decided to break up to different groups. They were called Task Forces. I recall initially there were 10 Task Forces. Among the 10 one was called GADS (Gateway Algorithm and Data Structure), and Dave suggested that I participate in GADS.
GADS held the first meeting in January 1985, but we only met a few times before GADS split up. From the beginning there were two kinds of competing agenda items. Some people were mostly interested in talking about blue-sky research, but some other people were more worried about burning issues in the operational network.
During the last meeting in 1985, it was decided to split the group into two: one was called Internet Engineering Task Force (IETF). The other one was called Internet Architecture Task Force (INARC). INARC met seperately for 2 or 3 years before it vanished. But the IETF kept growing like a snowball.
As I can tell, the only one of the original 10 Task Forces that still exists today is the end2end group, though it is no longer called a Task Force, but End-to-End Research Group, now under the Internet Research Task Force (IRTF).
Question: What kind of roles did you have during all these years at the IETF?
Lixia: Since the beginning I have always been part of the IETF, or more accurately the IETF has always been part of my life. In the early days I worked on routing issues. There was one closed WG under IETF: Open Routing WG. Routing is recognized as a fundamentally difficult problem since day one. Bob Hinden who was also at the first IETF meeting (see interview with Bob Hinden) chaired this Open Routing WG. We had around 10 or so people and met for a few years, trying to work out a new routing protocol for the Internet but didn’t get there. In the meantime BGP had been developed.
My thesis was on a different topic: today one would call it Quality of Service (QoS) across the Internet. I also did some work on transport protocols, mostly TCP performance studies. My first published paper (which got SIGCOMM Best Student Paper Award) illustrated some defects in the original TCP retransmission timer settings. Phil Karn invented a fix (about how to set the retransmission timer upon timeout) and that is the way TCP timers work today. I was also the first to observe the traffic dynamics resulted from TCP’s slowstart congestion control.
For some early IETF meetings I was the only woman and also the only graduate student. It was a very small research community, and was especially difficult for a foreign student to get in (Mike StJohns has some stories to tell about me here). I feel so lucky, being at the right place at the right time. All my life I have been lucky. Many people asked me how I got admitted to MIT. Well, God blessed me.
Around 1990 the ARPANET disappeared and the Internet became operational. That meant we did not have a playground anymore to test new things. So DARPA set up a small testbed, called DARTNET, and we used it to test out bunch of things, IP multicast, voice & video over IP, and also various packet scheduling algorithms. I recall at a scheduling algorithm bakeoff on the DARTNET, Andrew Heybey (a supporting staff for Dave) had to manually log into each router on the DARTNET to change the settings from one algorithm to another. Dave commented that “We need an automated Andrew”. So I worked out a signaling protocol, which later turned into RSVP.
Since late 90s I came back to work on routing again: looking at the global routing infrastructure, the dynamics, the scalability, and especially security.
Question: At some point you got nominated on to the IAB?
Lixia: The first time the IETF was still a relatively small community. I did a few pieces of work, my name got known. I was on the IAB from 1994 to 1996, only for one term, because I changed job and started teaching at UCLA in 1996.
Last year, I got nominated to the IAB again.
This time, the job on the IAB is fundamentally different than it was 10 – 12 years ago. Not only the community grew significantly, also the problem space has expanded enormously. There are more open issues, more challenges that we have to face.
Question: Do you mean the protocols and technical issues or also organisational and process issues.
Lixia: Both, the process and the vision. Where are we heading to? Fundamentally, what are the problems we are facing? In terms of process, but also in terms of solutions we are developing. Are we developing the right solutions? How do they work together? In the early days we used to know how every protocol works, and actually read almost all the RFCs. But pretty soon that was not possible anymore. The overall Internet has become such a complex structure that not one person can see all the pieces.
And I believe this is just nature.
I think there is a analogy here, even if it may be a remote one: When the network gets bigger, we get more applications, more users, more vendors, more operators, more protocols, more of everything. This is good, but it clearly makes the whole system more complex. In the old days we had three applications for a number of years: e-mail, ftp, telnet. Then the search engines started (archie, gopher). Then the web came and everything exploded.
I think there are two parts of complexity: when things get bigger, they necessarily have more parts. But there is another side of complexity: when a system is already huge, it is more difficul to design new components, to get it right. And lack of a complete understanding leads to overlapping and incompatible pieces, make the whole system more complex than necessary. I am afraid the latter is what we often see today.
As we developed more and more protocols, the network became very complex which makes it harder to understand. I believe we need a better way to see the overall picture insteadd of only focusing on individual parts of the network.
A big Internet also makes it more important to develop a vision. With a small system, if we were slightly off track, we would say ‘ooops, lets readjust’, and you could actually do it. But once the system is huge, it is much harder to adjust. Big systems have big inertia, that’s why it’s hard to turn. So I believe that today, more than ever, it is essential to have a clear vision. The network is this gigantic thing moving along that we have to steer it in the right direction, because any slight turn would take very long time and tremendous effort and cost.
That is one of my tasks on the IAB: to do my best to understand where the whole thing should be moving to. And how to nudge the community to move in that direction.
Finally I would like to make one more remark: at the first IETF I was a graduate student. I felt that I had so much to contribute. I got lots of great ideas. As years go by, I have better appreciation of how much I can learn from this community. Now, everytime I come to the IETF Meeting, I learn from others. That really changed totally for me: Now, I feel how little I know. As a graduate student, I felt how much I know .