The Light Reading Interview

In The Spotlight: Bernard Daines

The King of Gigabit Ethernet reckons he missed out on a $3 billion IPO by selling Packet Engines too early. Now he talks to Light Reading about:

In the past decade, Bernard Daines has played a key role in scaling Ethernet from its original 10 Mbit/s speed, to 100 Mbit/s and then to 1 Gbit/s. On one front, he's been a prominent architect of standards, and on another front, he's launched a series of startups that have rolled out switches implementing his ideas.

He sold his Fast Ethernet startup, Grand Junction Networks, to Cisco Systems Inc. (http://www.cisco.com) for $350 million in 1995. Then he founded a Gigabit switch vendor, Packet Engines, and sold it to Alcatel SA (http://www.alcatel.fr) for $325 million in 1998.

Now, Daines is back with a third startup, World Wide Packets (http://www.worldwidepackets.com), which was launched on January 1. Once again, Daines is thinking big. The new venture is making gear for Gigabit Ethernet access networks, aiming to drive costs to the point where carriers can offer 1,000 Mbit/s Ethernet connections to everyone, including homes and small busineses, for a mere $50 a month.

Light Reading: How did Gigabit Ethernet get started?

Bernard Daines: It started in the front room of my house in September 95. We issued a press release at the following Interop and put the paperwork together to take to the IEEE in November.

We [Packet Engines] actually shipped the first products. And because we licensed the MAC (media access control code) to so many other companies we almost single-handedly drove the interoperability of Gigabit early on in its life.

Packet Engines was a small, profitable company until we took funding in June of '96 to build Gigabit switches. We built the company up to about 225 people and then Alcatel bought us.

Light Reading: Did you want to sell Packet Engines?

Bernard Daines: I wish I could have held on against the VCs to go public the next year like everybody else did. It's what I wanted to do, but circumstances weren't right.

VCs always want a quick out. They see a bird in the hand or on the table they'll take it unless they're really, really convinced they've got the right team and everything in place to sustain an IPO. And of course, in the fall of '98 no-one knew that '99 was going to have so many multi-billion dollar IPOs.

I could have afforded to take the risk of saying no, we're going to hold out for an IPO and if it doesn't work, it doesn't work. But that wasn't the case with everybody else. The sentiment of the employees and the investors was here's a good deal. Dollar wise it was about as good as anybody had ever got at that time. I basically did what was good for the group.

It was a real saga on how to blow $3 billion in my view.

Light Reading: Why did Alcatel want to buy Packet Engines?

Bernard Daines: Alcatel was on a campaign to buy a company like ours, and one of the things they liked about us was that we weren't in Silicon Valley. Employee retention was very important to them, they said. Packet Engines' lead in rolling out Gigabit Ethernet in carrier networks was also important.

Of course, they trashed all of that in the way they handled things, but that was the original thought anyway.

Light Reading: Trashed?

Bernard Daines: Alcatel reneged on every promise they made to us.

We were turned over to a different crew when the acquisition was done. All the promises that had been made to us - of independence, doubling the size of the company, building us a certain building, leaving us alone, not sending hoards of people from France over to the U.S. to tell us what to do - all of those promises vanished within days.

The other thing that complicated things - besides the arrogant attitude of the French and not listening to anybody - was they bought Xylan. The people in Xylan are in some ways even worse.

You know Steve Kim [head of Xylan] is almost a mad-man. It's very common for his people to go around after him cleaning up the messes he makes. And Alcatel put him in charge.

Actually, even before the deal to buy Xylan was totally done, [Kim] came up to Spokane and went through the building saying this department's going, you're being fired, you're being let go. He stole the morale of the company and it's been going downhill ever since.

Light Reading: Where's the evidence of that?

Bernard Daines: About three engineers a week are quitting over at Alcatel Packet Engines. They find it almost impossible to recruit.

The other thing is that Xylan hasn't been able to manufacture the product [Alcatel Packet Engine's switch]. They have a lot of DOAs [dead on arrivals] on all their products and a lot of rejections. They're actually looking for a space to bring manufacturing back up here, after closing it down.

Light Reading: Gigabit Ethernet won quick acceptance in the enterprise, but slower acceptance among carriers. Why?

Bernard Daines: I failed miserably at the Sept '96 IEEE standards meeting that we hosted to try and get them to throw half duplex out of Gigabit There were purists in there that just couldn't bring themselves to do it. I pounded on the table, I pleaded, I cried, I begged, I said no one will build it, no one will ever use it.

I turned out to be right. Now people wished they hadn't put it in. It held Ethernet back from competing effectively with ATM.

When people got it, they started to put the QOS [quality of service] and COS [class of service] things into Ethernet gear. Now Ethernet provides a lot more flexibility in managing traffic than ATM, and QOS has turned out to be ATM's Achilles' heel.

In any case, all of this QOS stuff has gotten overblown. I'm more of a fat pipe person. Because of all the fiber and DWDM it's a lot cheaper to throw bandwidth at it than it is to manage it.

Light Reading: What's the strategy at World Wide Packets?

Bernard Daines: We're doing Gigabit Ethernet to the edge - the home, the small office, the strip mall, everybody.

Light Reading: Isn't Gigabit Ethernet overkill?

Bernard Daines: It makes working at home possible, even if you're doing drug modeling or ASIC design. It makes home education really possible. It makes HDTV [high definition TV] possible. Other access technologies - DSL, HSC, Coax, cable systems - won't do HDTV. They don't have the capacity.

Once you're looking at Ethernet, you might as well go for Gigabit. A 100 Megabit optical device for single mode fiber isn't that much less expensive than Gigabit one.

Light Reading: How are you financing World Wide Packets?

Bernard Daines: We haven't taken any VC money yet. We're going to raise our value so they don't get control of us. There are local investors that want to join with me as well.

Light Reading: Are you seeding any other startups?

Bernard Daines: I've invested in seven or eight small companies with supporting technology for what we're doing and headed towards. [Daines declined to give further details]

- Light Reading: Why TV over the Internet rather than cable?

Bernard Daines: The way the world is moving, there's going to be 2,000 channels for you to choose from very soon. Cable technology - using frequency division multiplexing to bring all the channels to your TV set so that you can select one with a rotary switch - isn't going to work with 2,000 channels.

People will select channels by looking to see what's on offer at different web sites. The Internet also provides the only way of distributing video from such a myriad of sources.

That's the reason why I totally disagree with people that think that circuit switching will make a come back for carrying video. It won't. Packet is the only way to go.

Light Reading: Where are carrier networks heading, then?

Bernard Daines: The WAN used to be thought of as expensive and scarce and hard to manage and error prone, but with all the fiber going in you can look at most modern structures as just being an extended LAN.

It will be Ethernet everywhere. It will be one big LAN and there won't be a need for anything else.

Light Reading: What role will Cisco play in transforming telecom?

Bernard Daines: They don't want this to happen, because they love to sell you something that's 15 percent better every two or three years and charge you all over for it again. They like gradual stuff so they can keep taxing you for the little increases.

Light Reading: Can you pull the rug out from under Cisco?

Bernard Daines:Cisco is now the second largest market cap company in the world - at least, it was when it hit $135 the other day. It's very difficult to pull the rug out from under them, but, hey, we all keep trying. It's fun to have a challenge.