A Net Evangelist Spreads the Word

This is the extended online version of the interview that appeared on page 26 of the October issue.

PSINet Inc. was established in 1990. With more than 21,000 corporate customers worldwide, PSINet is one of the world's largest and most experienced Internet service providers with services that include on-demand and dedicated Internet connectivity, remote access for mobile workers, corporate intranet and security services, and website hosting and development. PSINet Japan, the company's first international subsidiary, was established in October 1993. For more information about PSINet, go to http://www.psi.net/ or http://www.jp.psi.net/English/Welcome.html. Jeffrey Shapard is senior product manager (and "general Internet evangelist") for PSINet. For several years, until 1993, he was system operator of TWICS, the Tokyo-based online communications service.

interviewed by Wm. Auckerman

What new types of connectivity services will we see in the next couple of years?

Jeffrey Shapard: Most of the actual connectivity types of services are already out there. There's not a whole lot of really new stuff that is viable in the short term. Our energy has been focused on going back into the existing products that we have and beefing them up, expanding them, and adding enhanced features, rather than finding another type of connectivity product.

There's a lot of hype and hoohah about things like 56K modems and all the xDSL types of stuff, and cable TV/Internet. And there are various approaches to wireless networking. A lot of the excitement is about these esoteric datalink technologies that we have played a lot with in the lab, and experimented with, but these are not really viable in the short term.

Over the next few years, there is a lot of new stuff, new spectrum technologies, that will be interesting. But we've pretty much focused on the old tried-and-true, since we find that if we want to offer services in as broad an area as possible, we have to stick with the spectrum technologies that are available in as broad an area as possible. The more esoteric technologies tend to be very localized, only available in specific regions.

There are always articles in the press about how quickly the Internet is growing. Are those numbers really accurate?

Shapard: You'll read numbers like 500% Internet growth per year, but most of that is brought on by counting AOL and CompuServe and all of the consumer-type business.

The battle between online services and the Internet is over; the Internet won. Online has been subsumed into the Internet. All the serious online players are now in the Internet business in some form, even if it's only using the Internet to get access to their big centralized host computers. So when somebody like AOL says "OK, now we're an Internet provider," the Internet grows by 5 million consumer users overnight.

But we don't really count that market, because it's a market we're not in. We bailed out of the consumer market last spring Ewe had, at that point, 120,000 consumer customers, and we were losing money on every one of them. We realized that wasn't where our core strength is, so we sold off the customer base, closed down our New York operation, and laid off some people. It was kind of brutal, but it was a surgical strike. We had to do that as a company to regain our focus on what our core business is: the corporate Internet market.

The corporate market in the US is growing somewhere between 80% and 100% per year. Here in Japan, it's growing maybe 120% to 150% per year possibly closer to 200% today, though how long that will be sustained is hard to tell. That's the market growth. Our own sales growth pretty much tends to follow the market, or be faster than the market.

How is the corporate Internet market changing?

Shapard: The nature of the customers has started to change over the past year or so, from what we call the lead users, companies that really want to be out there on the leading edge with the latest technology, to more mainstream companies who are still a little bit nervous about this "Internet thing." They'd just as soon not have to deal with it, but they're increasingly aware that if they don't come up with their own internal Internet strategy if they don't start to use the Internet they will be left behind.

These customers are less interested in being a beta site for some esoteric technology, and more interested in basically having something that they know other people are using. They want something that they can be sure will work, that is tried-and-true and "bullet proof." We find that while there is a lot of excitement about T3 and xDSL, what most of our customers really want is a 64K leased line, or they want to start out with an ISDN-based, on-demand land connection. That's where most of the business is, so that's where we want to be.

Is an Internet presence a must for today's corporations?

Shapard: Definitely. By the end of the decade, any company that wants to stay in business will have to have Internet connectivity, will have to have e-mail, will have to have a Web page. Otherwise, they just won't be taken seriously.

How would you sum up PSINet's basic approach to the Internet market?

Shapard: Our whole approach to the Internet is: We don't believe that this is a data communications business. We're not in the data communications business, or the telecommunications business. The Internet is all about computer networking. As our CTO (chief technical officer) likes to say, the Internet is about shared business automation. The shared part is where the economics come in this big, shared network that brings a very interesting economic model in which the cost to each user is very low.

What people are actually doing with the Internet is sending e-mail, or exchanging files, or surfing the Web, or engaging in discussion groups. They're doing computer types of things, using it for different types of applications. The Internet is really about computer networking; the computers just happen to be distributed throughout the world.

In the end, it's an education issue. There are folks who actually come to us and say, "I want some of that there e-mail." Then when we ask about setting up an e-mail account for them, they say, "What! I have to use the Internet for that? You mean I need a computer?"

Where does Japan rank in terms of your overall company revenues?

Shapard: Japan is our oldest subsidiary; it's about 5% overall. We're about 600 people worldwide. We've got about 30 here in Japan, and that will probably double by the end of the year. In fact, I wouldn't be surprised if it doubles again by the end of next year.

From a market viewpoint, how does the potential of the Internet compare with that of corporate intranets?

Shapard: In the past year or so, there has been very strong interest in the use of Internet technology within corporate networks: the so-called "intranet boom." There's some speculation that there will be as much, if not more, growth in the corporate intranet segment as there is in the public Internet segment.

Actually, there's concern by some analysts that the "Internet market" will actually start to evolve in two different directions. One is public Internet connectivity and services, and the other is different types of corporate services based more on proprietary technologies although the reason companies are choosing Internet technology internally is to get away from the private, proprietary approach. Recent strategic flip-flops by companies like Novell and Lotus and Microsoft kind of indicate that they believe that too.

But there tends to be these two separate segments, with two very different kinds of customer bases emerging. There's some legitimate concern that they will take off in different directions. We're basically betting that they will follow somewhat parallel courses, and then will tend to merge again down the road, perhaps in a year or two when there are more robust security services tightly integrated into the technology itself.

What about the issue of security? Is it a major driving force for the market?

Shapard: It's a funny paradox. You'll read in the trade press about how security is the number one concern among corporate networking folks in getting their Internet connection in, about how that's really the barrier that has to be transcended. But when it comes down to how many corporations actually want to buy it, to pay the price for security the number is relatively small.

There's a big illusion that security is the number one concern, but we find that companies don't really want to pay for it when they see the price tag. They suddenly decide, "Hmmm, maybe it's not really that big a deal." What has been a challenge for us is to find security services that are really cost-effective enough so that everyone can use them. Solid security doesn't do our customers any good if it costs more than they are able or willing to pay for it. The ideal Internet security technology just is not available today, not from any of the vendors of fancy firewall packages.

I've seen conflicting definitions of just what constitutes an intranet. How do you define it?

Shapard: An intranet is the use of Internet technology for a private corporate network. Basically, it's another way to say "enterprise network," except an enterprise network can use any kind of protocol, whereas an intranet typically refers to the use of IP [Internet protoccol]-based applications.

A lot of the confusion about the term "intranet" has been driven by the software folks. They've taken a lot of different applications that happen to use IP, and relabeled it as intranet software. They've gotten a lot of mileage out of that.

What is an "extranet," and where does that fit into the picture?

Shapard: An extranet is what you call an intranet when you're late to market. [laughs] Seriously, as you start to get into more complex virtual business groupings if you're trying to create a private network connecting different organizations you get into more complex zones of security. "Extranet" is the buzzword that's used to refer to that. It's really an intranet with differing zones of access and security.

How would you characterize the Japanese Internet market?

Shapard: Japan has a supply glut; there are more providers than the market can really bear. But my feeling is that most of the providers aren't serious; they're not real. Every manufacturing company, every computer company, every telephone company seems to feel that they have to be in this business somehow, that they need to have an Internet service provider [ISP] just to understand what it's all about. So they take their WAN [wide area networking] group, and spin it off as an ISP. And they sign up maybe 20 or 30 corporate customers, most of them within their own business group, but that's about as far as they go.

The top 10 providers in Japan serve about 95% of the users. The top 3 providers probably serve about 50%. So there is a supply glut. But on the other hand, the market here really hasn't been scratched. There's enormous potential. Most companies in Japan don't even have LANs [local area networks] yet. They're just starting to crawl out of the 1950s and use computers. In terms of business automation, Japanese companies are somewhere in the late 1950s. Japanese efficiency in manufacturing is renowned, but in terms of actual efficiencies within the office, back office efficiency lags seriously. This isn't really unique to Japan, though; in Europe, it's a very similar situation.

But the future is wide open here. What I think we'll see in Japan is that rather than an evolution as in the US, it kind of goes in waves. Every time, there's a new step there's the minicomputer, so a bunch of companies buy that; then there are PCs, so people buy those; then a new generation of PCs, so a bunch of people buy them. When you look in a lot of Japanese companies, they have 20 years of IT technology inside their offices, with different workgroups having different generations of stuff, and they're trying to integrate it all together.

In Japan, companies tend to try to squeeze the total amount of investment they can out of this ancient stuff they've bought. Then they'll throw the whole thing away and jump into the latest technology they can get today. So instead of going through all these step-by-step layers, they'll just kind of leapfrog.

What does it take to become a big player in today's Internet market?

Shapard: To be a big player in the US, the barrier to entry is very high. There are very few companies that can enter the market in a big way anymore. There's AT&T: they've already entered the market several times, so they'll probably enter again at least a couple more times. They haven't gotten it right yet, but hey practice makes perfect. Maybe they'll figure it out someday. In general, the telephone companies haven't figured this business out.

What's ahead for the Internet?

Shapard: Who knows what it's going to be like in five years? We don't, and we've been doing this whole networking thing for a long time. But the Web caught us all by surprise.

And that's just the beginning. There's all this new stuff coming out. We don't know what's going to be happening even two years later; it's anybody's guess. Anyone who says they know what's going to happen in two years is either lying or they're drunk. It's a wild ride, and you just have to hang on, or you'll fall out.

What are the major driving factors of the Internet market?

Shapard: The three driving factors, or needs, underlying the purchase of Internet technology are first, to save money. Second, to make money a lot of which is pretty nebulous, like "let's put up a webpage about our stuff, so more people will see it and buy more of our stuff." And the third is competitiveness companies want to get Internet connectivity because their competitors are.

For mainstream businesses shoe manufacturers, or guys who make mufflers the Internet is a way to save money, particularly as an alternative to other types of wide-area networking. Typically, they'll go into it as a because it's cheaper than buying a leased-line WAN.

Companies realize that when they start to use the Internet for wide-area networking among their offices, what they've really done is created a common environment for all of their offices, and all these applications that they're using. Suddenly, a lot of them can do a lot more things internally than they could do before, and do them a lot more cheaply. Going out and getting a browser and public Web server is a lot cheaper than buying Lotus Notes, for example. And Lotus has realized this, and turned their stuff more into Web server-based.

So, companies initially go in to save money. But it opens up a whole new realm of possibilities that help them increase their productivity.

Coming back to the issue of security, is that the greatest area of concern for corporations on the Internet?

Shapard: There's this image of the Internet as a wild and woolly place, that if you're out on "the Internet," then your data or your secrets can be compromised. But there is no "the Internet" out there. It's a whole bunch of providers, guys like GOL, IIJ, MCI.... It's the phone companies. It's a whole bunch of different networks that somehow will talk with each other. There's no "the Internet" backbone.

The big guys have pretty secure networks. Our network is as secure as the phone company's network; actually, access into our network is more secure than the phone company's network. It's really a matter of helping our customers, and the market, become more comfortable with the new generation of service provider. You never think twice about questioning the phone companies.

We have huge respect for the privacy of our customers. We don't even do monitoring of traffic. We only look at things like volume, or the load on a particular link things that don't require us to open up people's packets. Most ISPs will monitor traffic to gather statistics, to see what percentage is telnet, what percentage is ftp. et cetera. They'll monitor the different kinds of traffic but to do that you actually have to look inside of people's packets.

To us, looking inside of someone's packet, even just for the network status, is a violation of privacy. So we don't even keep those kinds of stats. Because if we did, we'd have to have tools to allow us to look inside of packets, and we don't want those tools. And if we have those tools, then maybe along comes some police agency and says, "why don't you guys help us, and kind of watch for a keyword or two." We just don't want to do that.

What are the key elements of ensuring Internet security?

Shapard: Often, Internet security is discussed as if it is this unique thing. But security is security. You should have a similar type of policy for governing security on your LAN as you would have for governing security for your building And about 95% of it is about access control. That's the biggest concern for Internet security, because it's the biggest concern for security policy in general.

The transit issue, data going across the network, has more to do with privacy and confidentiality can they see who I'm talking to, can they see what I'm saying? That's a big concern, but there are various ways to address that. Encryption is one of the more popular mechanisms, but it's also one of the least efficient mechanisms. And authentication is a concern that's often overlooked how do you prove you're who you say you are? The there's integrity, which has to do with whether somebody can corrupt your data, if what you've sent is what is actually received.

The more security you implement, the more of a pain in the butt it is. That's just part of security. The biggest amount of IT crime is actually done inside the office by employees of the company. And it's vastly under-reported. The problem is more often a disgruntled employee ripping off the company than it is some wily cyberterrorist member of the CHAOS computer club. The Internet security threat is overblown. Companies should be aware of it, and secure themselves against it, but the idea that the Internet is somehow inherently more insecure than their own office is definitely an illusion.

What's the future of Internet telephony?

Shapard: The best network architecture in the world in terms of efficiency, in terms of actual cost for the provider per call, was designed 80 years ago. The telephone network is the most efficient network in the world for telephony. The Internet is the least efficient network in the world for telephony.

The actual cost per call for using the Internet for telephony is 10 to 50 times higher than it is for the regular telephone network. Now, the actual price you pay is more a matter of regulatory structure, universal access requirements, monopoly, industry structures, this kind of thing, than it is about actual cost.

So, the idea that the Internet is going to take over the telephone system because somehow it's free bandwidth.... The Internet is not free bandwidth. It's very expensive bandwidth! Guys like us, we have to go to the telephone companies and lease circuits from them, and they make a profit. They're making money hand over foot just leasing lines to guys like us for our customers and our own network. They make a profit every time they lease a line to us, and then we add in our switches, our routers, our support groups, and all that....

So, our bandwidth is actually pretty damn expensive. It's just that we don't charge by the minute, or by the bit, so it looks like it's free. It's not free!

What's happening in this rush to the Internet for telephony is that basically people want a free ride on our network. But we're not really interested in getting into a business where we're not going to make money.

Some people say, "Oh, it's all IP; it works just like normal Internet applications." Well, it doesn't, because most Internet applications are running on top of TCP. And TCP is based on this idea that you will send a bunch of little packets, and if some have to wait for awhile, they're going to get through eventually; but it also has a bunch of error checking in it, because it wants to make sure that every packet gets through. If it doesn't get through, it's going to it's send it again.

When you start to get into telephony and video, you start to get into more streaming technologies where it doesn't really matter if every packet gets through, just as long as enough get through. If they all have to be checked, it will just totally glom up, because some will surely get lost, and it will wait until those get resent, which brings everything to a halt. So you end up running on top of UDP, which is a kind of lame protocol, but it's fast. It just throws packets across and doesn't really care if they all get through.

If there's a lot of competition on the network, hey, maybe your e-mail takes a second instead of a half-second to get through. Fine, but that can kill a telephone connection.

Are you going to care if it takes 300 milliseconds or 800 milliseconds for your e-mail to get through? Can you even measure the difference? As long as you send it, and it gets through in a reasonable amount of time which might be a few seconds to a few minutes you're a happy camper. But if you're trying to do a phone call, and it keeps breaking up and losing the connection, you're not a happy camper.

So, suddenly, there will be different tiers of service. To be able to guarantee the quality of service that's needed for business applications, which is really where you make all your money in telephony, you have to offer a certain quality of service. In other words, you have to start doing bandwidth guarantees. And you know what? When you start having to do bandwidth guarantees with Internet service, it starts to take on a very different economic structure one which looks very, very much like the telephony model.

Bingo. So much for the myth of free bandwidth across the Internet. If you're using [Internet telephony] over networks that aren't that busy, the voice quality is all right. But when the providers its riding on top of catch on to it, they'll start charging Internet telephony service providers real money. And it won't be viable economically.

Will shortage of bandwidth ever be a problem for the Internet?

Shapard: No! The Internet isn't about bandwidth. Bandwidth is not really very relevant. One of the big myths of the Internet is that bandwidth is somehow some kind of big constraint.

Basically, the determining factor of Internet performance isn't bandwidth; it's packet loss. The only time bandwidth is really the limiting factor is if you have what we call "chronic packet loss" when you have a small provider with a 64K link, and 2,000 customers all hitting porno sites , in which case the connection is totally maxed out and they're just losing packets left and right because of bandwidth constraint. But when you start getting into T3s, typically they'll be running at maybe 15% capacity... to have room for the "burst" transmissions.

So, if the solution isn't more bandwidth, what's the key to being a successful ISP?

Shapard: The Internet is more about managing routers and the topology of the network and the queues, the buffers, the switches than it is about bandwidth. And I think that's a pretty important message, because there's this myth that somehow it's all about bandwidth: that the bigger the pipe, the better. If you look at some of the providers who over the past few months have had the most significant performance problems and network reliability problems, they're telephone companies. They have all the bandwidth in the world, and yet their networks keep crashing. It's not because of the bandwidth; it's because of the routers, and inadequate management of buffers.

The folks who believe that the Internet is about bandwidth should get out of the business immediately. They're going to lose it. If they're right, we're all dead, because the telephone companies can win that game hands down.

The Internet is about IP. It's about shared business automation; it's about e-mail, and Web services, and groupware. It's something different than telephone, because the telephone network is beautiful in its economic and technical elegance. It's just tariffed in a goofy way.

A final question: What should the average corporate user look for in an ISP?

Shapard: You should look for experience.

Most small and medium-size companies are not going to have an Internet expert in their company. They may not even have somebody who knows how to back up the file server; they may have a systems integrator who comes in sometimes to do that. So they really need to find somebody who's got experience, and that they're comfortable with to really to take care of what will become a very important component of their business processes.

Another consideration is to look for somebody who seems like they're committed to the Internet. There are maybe 150 providers, mostly in Tokyo, who offer business services. But most of them are companies who've got into this within the past year or so; they're companies whose core business has nothing to do with the Internet. It's just like a sideline they've stumbled into, and it's not even clear that they'll be around a year later.

Another thing to look at is your growth path. Companies will typically buy less than they probably need. So if you come into a company that doesn't really have a rich set of services, then you're going to be out shopping again in six months to a year. But if you come into a company that essentially has a growth path built into it, there's room to grow.

When you're evaluating an ISP, find out how much they really understand IP. If they basically talk about bandwidth and data communications, turn around and run away, because they don't know anything about the Internet.