Following the news that the Metro Ethernet Forum (MEF), has launched new specifications for metro Ethernet designed to enhance interoperability of services across vendor offerings and so open up competition, we interviewed Bob Metcalfe, the inventor of Ethernet.
Q: What was the deciding factor in halting development of the original basic Ethernet idea -- in other words, at what point in the design process did you decide that it was good enough?
A. We never actually halted the development of original Ethernet -- but when Dave Boggs and I were building the first wire-wrapped card, when the board got full we stopped. The CRC chip was the last thing to go on and even that was hanging off the edge of the board.
Q: You're famed for having predicted that ATM would overtake Ethernet -- and then to have publicly retracted that. What was the biggest issue which persuaded you that ATM was going to win -- and was it economic or technological?
A: When I became a pundit and was practising journalism in the 90s, I decided to write a story abut ATM. I interviewed 32 top people and they all said that because of QoS, other complex technologies and the huge level of investment behind it, that ATM would invade the LAN where Ethernet was king. My friend George Gilder was aghast and said I was wrong. He was right.
Q: Would you go back to 3Com, if they asked nicely? If not, why not?
A: They're now based in Marlborough Massachusetts, just down the road from where I now live so if they did, I can't say anything except maybe -- I'm not sure. I haven't been there since 1990 though and I'm not sure I'd be of any value - I've moved on to other things since.
Q: Ethernet is threatening to move from a local to a wide area technology. Did you ever foresee this -- if not, why not?
A: No, I didn't. There's a distinction underlying your question, that is, between a service and the technology. The service is the interface, the technology is the stuff underneath. And it's the interface - the service - that the Metro Ethernet Forum [the IT industry consortium promoting Ethernet in the metropolitan area] is presenting. This can be done over SONET, ATM or Ethernet. In other words, a T1 is a service interface. You can carry T1 service over Ethernet technology or vice versa. The MEF is discussing Ethernet more as a service.
Q: Do you have any advice for budding engineers out there?
A: Yes - I'm very enthusiastic about engineering and my advice would be that it's about developing disciplines and tools for controlling complexity. Our systems are very complex -- too complex -- and the best way of managing that is to develop frameworks of abstraction to control them. Systems you build are never debugged totally so they need to always to carry debugging tools. Something will always go wrong. You can't predict what nature, users or bugs will do - bugs will always be there.
Q: You once said that staying at 3Com was the hardest thing you did, after they fired you as CEO. Is it still the case -- and if so, what comes closest?
A: To get my Harvard PHD after they rejected my thesis at the eleventh hour in 1972. The lesson is that I was insensitive to politics. I hadn't paid enough attention to the world at Harvard because I was employed at MIT and wasn't connected enough to know that I was going to fail. Eventually did get the PhD thanks to Xerox who asked me to come on out to them anyway, even though the employment was contingent on the PhD.
About Ethernet generally
Q: What's the biggest technical factor which has made Ethernet the success it is?
A: It's about several things. Firstly, Ether-centric means being distributed, unlike the phone network where the intelligence is in the centre and the phones are dumb. Ethernet is the opposite, the CPE is king not the network. Secondly, the network is packet-based. Next, Ethernet is fast. The original 2.94Mbit/s - that's still faster than a T1/E1 runs now. Then, Ethernet is based on a layering of protocols which makes it simple. It understands its place in the layer hierarchy and doesn't try to replicate functions of higher layers. And finally, it became open and as a result, the dominant standard. Others made the mistake of trying to be proprietary and it didn't work.
Q: What lessons we could learn from Ethernet's having been 'good enough' networking rather than the most technically advanced -- such as ATM?
A: ATM isn't advanced - it's just complicated. It solved the wrong problem. Ethernet understood its place and was simple, totally unlike ATM.
Q: Which other networking technologies would benefit from a dose of 'good enough'?
A: Dunno. 802.11 & .16 - maybe, but I'm not in deep enough to be sure.
Q: Can Ethernet get any easier than it is -- and if so how?
A: Wireless makes it easier, IPv6 makes it easier but I can't think of anything else.
Q: Are there any limits to the speed that Ethernet can reach?
A: No. The original [collision-based] CSMA/CD technology won't go far because the detection of a packet collision to get from one end of the network to the other, so you're limited by the speed of light. We've moved beyond that now we don't use CSMA/CD.
Q: People talk about Ethernet to the home. How long do you think it will take before it becomes a majority method of Internet access in the developed world?
A: Five to 10 years, though it depends how you define Ethernet. I claim it's Ethernet access if the data being carried over the wires is in Ethernet packets. In that case, it's Ethernet.
Q: Ethernet has had 30 successful years. Can you foresee anything changing that in the near to mid-term future?
A: Last year were 184 million new Ethernet ports shipped, according to IDC and it appears to be accelerating. So in spite of changing, the Ethernet philosophy is continuing to proliferate and I'm grateful that the Ethernet name has stuck to that, unlike the wireless technologies like Wi-Fi and WiMax.
About wide-area Ethernet
Q: Do you see a clear way through for Ethernet to become the wide-area networking standard? In other words, what are your preferred technology or technologies that will make it all happen?
A: That's what's happening now - we're proliferating Ethernet as a service interface - it'll be an Ethernet service which is what the MEF is promoting.
Q: What are the main barriers for metro Ethernet today?
A: Three things: the effort involved in migration because there's a lot of infrastructure and migration is slow. Secondly, Ethernet is only now being re-specified for use in the metro area. The MEF's announcement of two new specifications preparing Metro Ethernet for metro area use. Lastly, the carriers are monopolies so we need to increase the competition -- thank god for the cable companies which will accelerate proliferation of Ethernet.
Q: Is metro Ethernet becoming too complex or too remote from the original concept to be called Ethernet? If so, which and why? Is it still Ethernet as we know it?
A: Ethernet today - there's very little resemblance to the enterprise Dave and I built in 1973. It's evolving and the name has stuck until now and it's hotter now than it's ever been. I'm grateful since Ethernet is my word. But it's a marketing question and right now the market is voting for Ethernet.
Q: Metro Ethernet is seen as being a huge cash cow for operators selling services as opposed to CPE vendors, even though operators have not been very good at selling services so far. Can that change?
A: Operators are cash cows and that what Metro Ethernet hopes to change - we want services to be better, faster, cheaper, not monopolies. Metro Ethernet is aimed at packet-oriented services using competitive technology and that will drive prices down.
Q: The Metro Ethernet industry divides into two broad camps. One is focused on pure Ethernet, the other on updating and adapting SONET networks. How can the two approaches be reconciled?
A: Because of protocol layering, SONET services can be provided over Ethernet transport and vice versa, which creates opportunity for migration to an Ethernet transport, and allows operators to move to the new technology at a sensible rate.
BT recently announced it is going to convert its entire voice network to IP. Will other operators be forced to follow?
A: Yes. My quibble with the question is that most are going in that direction. The world -- all the other carriers - is doing it anyway.
Do you think that businesses and carriers yet understand clearly what metro Ethernet is - and if not, what can be done to improve matters?
A: It's a gradual thing - it's what the MEF is all about, promoting interoperability between services and vendor technologies. Since we're going to be carrying all types of data including voice, it's appropriate that we move from a voice to a packet based hierarchy. It's not going happen suddenly.