/

Internet Economics

Author Hal Varian sounds off on issues of the 'Net.

Hal Varian is the dean of the School of Information Management and Systems at the University of California at Berkeley, and co-author (with Carl Shapiro) of Information Rules: A Strategic Guide to the Network Economy. Network World Editor in Chief John Gallant recently talked with Varian about the problems inherent in the current peering arrangements among Internet backbone providers and other economic and business issues shaping the Internet.

Q: Can you talk about the economics of the interconnections among the backbone providers today? Based on my reading of your work, it seems you don't think the current system can last much longer.
A: [The current system] will have to evolve to a more stable business model. The thing that will drive it to that is consolidation. What is happening with the backbone providers is the establishment of large economies of scale. There are maybe 20 providers today, with six big ones, and we'll wind up with four major players if the WorldCom/MCI merger goes through, as it looks like it will. We're going to end up seeing a handful of backbone providers - no more than six and probably four - of major players in that industry, just like we have three major long-distance carriers. In some sense, in purely technological grounds, you could get away with even fewer, but I think antitrust and regulatory issues will stabilize the number around four.

Q: Wouldn't a smaller number of backbone providers actually make it easier for the providers to handle interconnection on an informal basis rather than to have some formal mechanisms?
A: Bilateral interconnect is relatively stable from an economic point of view. But what will change this is quality-of-service considerations. You have to provision your network differently if you are sending different shapes of traffic. If I am sending you a lot of real-time video feeds, that's one thing. If I am sending you store-and-forward e-mail, that's another thing. QoS issues will push pricing in general and when you have pricing in general, you will have to have some kind of settlement method.
I could be wrong. There are people who vehemently disagree with this issue: they say, 'we don't want settlements. That's just a bunch of overhead that isn't worth the trouble.' On the other hand, show me an another example of an industry where you've got this kind of gentleman's agreement for handing off traffic, especially when the participants are so symmetric. It is certainly not the case in the telephone network, absolutely not.
Go back and look at the historical record on this. Look at how the Bell system achieved its prominence in the 1920s. It was by strategic interconnection policies. Bell invested in long-distance early on and then as the demand for long distance grew, they said to local telephone companies: 'Sure, you can have access to our long-distance network. Just become a Bell affiliate. Join the company.' They wouldn't allow interconnection to that system until the locals became affiliates. The same thing happened with Marconi and wireless. Marconi absolutely refused to interconnect with any other wireless systems. That is how they maintained their dominant position for 25 or 30 years.

Q: Have you seen any signs in the current market that companies are starting to abuse interconnection because of their market dominance?
A: There are two senses of abuse. One sense is so-called 'hot-potato routing' where the idea is that since you are paid by the end customer and you have a flat interconnect agreement, your incentive is to get rid of the traffic as soon as possible and bounce it off to someone else. So that is a distortion created by the current mechanism. That wouldn't happen if you had traffic-sensitive pricing. On the other side, the question of whether firms have tried to exploit their monopoly power, I'd say that up to this point it has been a relatively competitive environment. When UUNET tried to switch from peering to customer agreements last year they got a lot of heat and backed off.

Q: Was UUNET's move a harbinger of things to come?
A: Yes.

Q: So going back to this QoS issue, your sense is that as companies begin to offer service with QoS guarantees, they will expect more out of interconnections and this one-to-one peering won't work as well.
A: Right. Here's what happens: I want to do a videoconference with a customer who is on another network, so we have to send this video stream across this interconnection. The first carrier gets paid for providing that service to the company generating the video stream. It will want high QoS on the other side, the network receiving the video stream. But they're not getting anything out of this. So there has to be some transaction between those two networks that says 'I'm going to treat this traffic in a certain way. I'm going to give it a higher quality of service.' That is more costly, you are imposing extra cost on that company, so there will have to be some kind of compensation to make them willing to bear those extra costs and provide that QoS. That necessitates a transfer of money. Now, maybe it is not worth it if you are talking about some delayable low-quality service that isn't worth tracking. You just provide a generic level of service, for e-mail, say. But as soon as you have quality differentials, you start requiring pricing, and when you require pricing you need to have transfers at the interconnect points.

Q: What are the key elements of what you would call a "fair and reasonable" interconnection agreement?
A: Fair, reasonable and nondiscriminatory is the classic phrase that is lifted out of licensing agreements. When two companies decide to share some intellectual property, do a cross-licensing agreement or get together to set an industry standard, they will agree to license their technology on a fair, reasonable and nondiscriminatory basis. The amazing thing is that that works. Every now and then there is a case where someone complains and they go into arbitration and they end up with a resolution. The intent of my choosing those words was that you want something loose enough so that it can be flexible, to respond to changing technology and changing conditions, but also have some court cases behind it and some legal opinion that would allow you to reach a resolution in disputes. In my view, the emphasis should be on the nondiscriminatory side. Suppose the WorldCom/MCI deal goes through - though now it looks like the MCI part is being spun out of that deal - suppose you controlled 60% of the backbone traffic. Well, you still have to interconnect with people for the other 40% of the traffic. That's valuable to you as a provider, so you are looking for attractive terms for those interconnects and if you have attractive terms with those interconnects you have to interconnect with new entrants like Level 3 or Qwest on the same terms. If you have nondiscriminatory as the operative feature, you can go a long way.

Q: Do you mean terms that everyone in the industry agrees to, a common set of interconnection terms? Or do you mean that one company must use consistent terms in all its dealings with other carriers?
A: I mean one company using consistent terms. First of all, you wouldn't want an industry standard that would be imposed in any way because it would be very hard to evolve. You have rapidly changing technology and you have to have flexibility in these terms. One of the things that I think would be very bad would be if some regulatory body, like the FCC, got in and starting setting terms because you would end up with very inflexible conditions. That's why I suggest that you have an industrywide arbitration board.

Q: Is that the problem we see with local telephone service, that the regulators have tried to come up with a single set of interconnection standards?
A: Let me give you a prime example, just a perfect example of this. We had a slugfest out here a few months ago between Covad and some of the RBOCs over the resale of the local loop for high-speed data. The RBOCs maintain, and I think this is correct, that there is a big difference between data-quality copper and voice-quality copper. You can have a local loop that is perfectly adequate for voice but if you want to send ADSL, you have to go out and condition it. You have to close off some unterminated connections, you might have to upgrade some of the equipment on the line. That is expensive. Any time a truck rolls you are looking at $500. Covad says you have to sell this line at $12 per month, or whatever the mandated rate is in California, but the RBOCs say it costs $500 or $1,000 to get that line ready for data. We're willing to sell you voice-quality copper for that price, but not data-quality copper. What happens when you get that kind of dispute? Look at the telecommunications law of 1996 and of course it is absolutely mute on this issue because they never thought of ADSL. They never thought you would be able to use this line for data. They were thinking almost entirely of voice. You are having a big delay in the deployment of this technology as these groups slug it out in front of PUCs and other regulatory bodies to try to reach some resolution. That's the problem: we aren't going to be able to specify what the technology will be two years down the road. Anything that we write into law and bureaucratic procedures now will likely end up being a drag on new technology. That is why you need a faster means of arbitration and dispute resolution in this industry. The difficult part of network industries is that everyone depends on everyone else, so you have great complementarities among different players. The good side of that is that if you don't play along and agree to behave appropriately you could be shunted out and left in a difficult position. Arbitration and other more informal kinds of dispute resolution can be quite effective in these industries.

Q: As devil's advocate, though, without a top-down approach, say a federal regulatory approach to interconnection, couldn't a company like WorldCom/MCI, as originally envisioned, create consistent terms, but still abusive terms for interconnection?
A: If you are talking about a company with 90% market share, maybe. If you are talking about 60% market share, that seems unlikely. You've got to connect with that other 40%. There is a lot of the country that MCI/UUNET and MFS wouldn't serve, so they would have to get connectivity to those parts of the country in some manner. That requires interconnect with the other players.
It's like we've seen with Microsoft. Nobody worried when Microsoft had a 60% market share, now they have a 90% market share and everybody's worried - for the same sorts of reasons, interconnection. Interconnection with other software providers and with other providers of complementary services.

Q: How does this play out internationally?
A: The international situation is so in flux with some countries still having governmental telecommunications, many are in the process of privatization but they still have a dominant carrier left over from the government provisioning days. It is a situation where you really have to look in detail at each country. I don't think there is a general, blanket statement of what will work.

Q: Who should organize the kind of arbitration board? Should it be the leading players?
A: You need some governmental urging to make this happen. I don't think you will see a lot of progress until you see more industry consolidation because once you get away from the top backbone providers, there are still 3,000 to 4,000 ISPs and at least 6,000 different opinions of what should be done. You would need some government urging to make it happen because it won't happen on its own.

Q: Should the IETF or an organization like that spearhead it?
A: I don't see this as the strong suit of the IETF. They are much more capable on technology issues. On this legal/social side, they are not the best player. The Internet Society is kind of weak. Part of the problem is that there is a huge vacuum in terms of these issues of Internet governance. I was kind of hoping for something to come out of Washington with this domain name system dispute, but Washington ended up punting. They knew they punted.

Q: You recommended that the Department of Justice push for this kind of fair, reasonable and non-discriminatory interconnection as a quid pro quo for approving the MCI/WorldCom merger. Do you have any sense that the department has explored that opportunity and may do it?
A: I spoke at the FCC two months ago and talked to the people working on this issue. I brought this up as a possibility. They all said, "Hmmm, interesting idea. We've done things like that before.' That's the extent of my knowledge. This is not the kind of thing the DOJ has typically done because they usually focus more on spinning off pieces of the business or certain conditions at one point in time that are imposed to deal with an anticompetitive threat. It is more like something the FCC would do. The nice part is that there is always this threat of regulation if you don't do it yourself. It's much like the debate going on about industry self-regulation in regards to privacy. What is happening is that the government is saying that the industry has to get its act together with respect to privacy and if you don't do it, we'll do it for you. I think you need that big stick back there to make this work effectively.
Again, look at the early history of radio - that is how the government actually broke the Marconi monopoly. It was war. World War I put it under a lot of stress so they required them to interconnect and by World War II it was pretty much gone. So if we had a war we'd be in great shape.

Q: Let me ask you about a different level of interconnection, that between the ISPs and the backbone providers. Does that system work today or does it need to have more formal guidance?
Q: The trouble is that the term ISP spans a large range of business sizes and practices. On the one hand, there are these small ISPs that are basically providing technical support to customers and are outsourcing the actual interconnection. That model will become more prevalent in the future. You will have network support businesses that work as intermediaries between providers of transport and customers. From that point of view, if you have a competitive environment with lots of different providers of transport capability, then that is good. You want to deal with a competitive market because you generally get low prices and better services. There is no real need to worry so much about the terms of those contracts. What is important to worry about is making sure the market is reasonably competitive. That means you want to make entry into that market easy. I suppose when you talk to the Level 3 people and Qwest, one of their worries is how they interconnect with UUNET - what are the terms, how do they get into this business. That is where my first proposal is the way to go - namely, if you had an arbitration procedure and open, explicit terms of interconnect - you could carry that out.

Q: Jim Crowe of Level 3 has kicked around the idea of IP equal access, where the industry would come together to spell out the terms of interconnection - what services do I buy, what do they cost and how are they provisioned.
A: Let me tell you the down side of IP equal access. The down side is that if you do get three or four major players that is a really nice way to run a cartel, isn't it? You all agree on uniform, standardized terms and conditions and pricing for newcomers. You don't want uniformity across the industry. That would be a bad idea. Too easy to abuse. But you should have uniformity within companies.

Q: As long as UUNET's terms are consistent across the board, that's acceptable, but you don't need to do that on an industry basis?
A: Right. And I would go even further and say it is dangerous to do that because of this cartelization problem. I can tell you that the DOJ would not look favorably on that kind of arrangement.

Q: Even if this arbitration board represented all aspects of industry?
A: Standardizing technical aspects of interconnection is great, that's fine. As soon as you start talking about pricing, terms and conditions, etc., then you will get into trouble.

Q: Well, how would it be different from interconnection at the local level where we know what unbundled network elements cost and what the conditions of interconnection are?
A: Remember that is a regulated industry we are trying to move to a competitive industry. This is the opposite situation. We've got independent operators who are competitive and they're somewhat worried about concentration in that industry now. If you had an industry group which tried to set prices, that would be a huge red flag in terms of cartelization. That just wouldn't happen.

Q: What's required to ensure that the ISP business stays vibrant?
A: Again, to use a historical analogy, it's a lot like the situation with the small markets and the supermarkets entering in the 1920s and 30s. You had these small local merchants and the big supermarkets came in and there were a lot of calls for legal action to prevent that kind of competition. That's where we got the Robinson-Patman Act, which is probably the worst antitrust law ever passed. We see the same thing with Wal-Mart and in Japan. The industry is evolving and changing. We will still have choice in the future, but you will see more consolidation. Some companies that are ISPs now are going to evolve into network service integrators, to coin a phrase. They will basically be doing the integration with the larger companies who are providing the services more efficiently.

Q: If the industry doesn't take any steps toward creating this arbitration board or iron out the guidelines on interconnection, how long do you think the current peering system can last?
A: The crunch will come, the situation will come to a head, as Qwest and Level 3 come online. They've got a lot of capacity, they want to provide IP services and they're only going to be effective as competitors if they can interconnect with the other large incumbents. There will be a lot of teeth gnashing and disputes on the terms of that interconnection. I'm sure there will be legal cases filed. I'm not sure what language will be used, but that will be a struggle.

Q: Who should be represented on the arbitration panel?
A: If you get explicit government representation, there will be a lot of rules that go along with that. That's bad. So you probably don't want explicit government representation. But there is a worry that the group could be too industry-dependent. You'll wind up with a capture problem - the cartel situation. You want neutral parties, some academics, some people who have previously served in governmental or industry positions - people like Vint Cerf. The only problem is that he is with MCI now, but that flavor of people. People who have established a reputation for public interest. Also IETF people, that sort of crowd. You will also need people who can deal with arbitration, experienced people in resolving these sorts of issues. The difficulty is that when these things end up in court then you end up with a random draw of judges who may not what TCP/IP stands for or who have no idea. You need people who understand enough of the technology and enough of the legal and institutional side of things that they can make effective decisions.