A few years ago, then-FCC Chairman Michael Powell tried to justify his rolling back key pro-competition provisions of the Telecom Act of 1996 by stating that they were no longer necessary. Competitive providers didn’t need access to incumbent telephone company facilities because they could compete via “intermodal competition.” So while the existing modes of voice-band and “broadband” access — the telephone company’s network and the cable TV network — might not be available to competitors, technology would allow them to find alternatives. For instance, they might be able to make use of microwave radio (except for the pesky matter of the FCC’s strict licensing rules), or of power lines (a ridiculous idea), or invent something — perhaps he meant subspace communications, like on Star Trek. Yeah, that’s the ticket — make regulatory policy based on unavailable or not-yet-invented options, or science fiction. This fit in perfectly with the White House insiders’ preference of faith, versus conceding to the “reality based community.”
But while intermodal competition was the cover story, the real story was one letter away. The incumbent telephone companies, with their patrons at the FCC, have been fighting off a much bigger threat, that of intermodel competition. It’s a “battle royale” between the regulatory and business model that underlies the telephone industry and the very different model that has sprung up to support the Internet.
Born of Monopoly
The telephone industry was actually quite competitive for a brief time, after the expiration of Bell’s patents in 1893. But AT&T (News - Alert) had a number of tricks up its sleeve and within two decades, most of its competition was gone. It offered instead to be a regulated monopoly, with a few thousand other little telephone monopolies serving the mostly-rural areas that AT&T didn’t consider worthwhile. Absent competition, prices were divorced from cost. Instead, prices were based on “value,” as perceived by regulators. Rural carriers were subsidized by urban customers, and an arbitrary boundary was drawn between low-priced “local” calls and high-priced “long distance.” The fixed price of access lines was kept artificially low. This probably encouraged “universal service” back in the days when most people couldn’t otherwise afford a telephone. It also gave politician/regulators a simple regulatory benchmark: They were doing their job if the monthly fixed residential bill, known as the 1FR rate, was kept down.
As competition was introduced into the industry beginning in the late 1960s, new barriers were created to protect that arbitrary distinction and its associated subsidy flows. So, long-distance companies pay access charges to local phone companies at each end of a call. These rates can be quite high: While the average rate for both ends of an interstate phone call between Bell companies is about 1.2 cents per minute, rates for some intrastate calls can be as high as 15 cents. It could actually be cheaper nowadays in many states if somebody set up a “call back” service in Canada or even China to connect two intrastate points at today’s low international wholesale rates! Of course this would put state regulators and the FCC in the uncomfortable position of opposing a strategy which they encouraged a decade ago as a means of lowering international call termination rates.
One of the reasons voice over IP (VoIP
) is so popular, of course, is that it allows avoidance of paying these rates. A PC-to-PC service like Skype never touches the telephone network, while Vonage (News - Alert) and its many imitators take advantage of an ambiguity in the rules (the FCC has been mulling over the proper treatment of VoIP for about a decade now, and is still undecided) that usually allows these access charges to be avoided.
And that’s where the threat gets real. It’s not that Internet Protocol is always more efficient than traditional time division multiplexing (TDM
) telephone technology, especially for carrying voice. Occasionally it is; often it’s not. But it’s the business model that’s different. The Internet’s business model doesn’t charge users for distance and it usually doesn’t have any kind of visible usage-related charges at all. You pay for a connection of a certain size, but you don’t get any guarantees of end-to-end performance.
The public Internet doesn’t have room for cross-subsidies taken from monopoly profits: It was born a competitive market, and its prices reflect the impact of competition.
Why Do People Like VoIP?
The growing popularity of Vonage and similar VoIP services certainly isn’t a result of call quality. How many people really want their landline calls to have the dropouts and latency of “bad cellular,” which is what a lot of VoIP still emulates? Do they really want their phone calls to fade out when uploading a file or hitting the “send” button on e-mail? Nor is it a result of its reliability, which is still not up to most telephone standards.
What drives VoIP is the perception that it is not The Phone Company. Instead it brings the Internet business model, wherein voice is seen as being like other applications, where distance and political boundaries don’t matter.
If the same business model could be applied using PSTN technology, it might succeed there too. But the monthly rates for users who receive subsidized phone service would go up, and that would upset politicians, especially farm-state senators. Investment companies who have gambled on subsidized rural ILECs, including politically-connected ones like Carlyle Group (Hawaiian Telephone) and Goldman Sachs (Madison River), would also resist. So the PSTN will be left to gradually decline, its subsidies (including the USF tax) rising in a death spiral as unsubsidized users abandon it.
Wall Street has however harped onto VoIP as a technology, often not realizing that IP is an artifact, not the cause, of the Internet’s success. It’s a secret sauce syllogism: The Internet is good; it uses IP; IP therefore must be the secret sauce that makes it good. But coincidence does not demonstrate causality. Maybe IP technology is good enough to support the Internet’s business model, and that’s the real secret sauce.
During the Internet bubble, Lucent Technologies (News - Alert) introduced a primitive VoIP upgrade for its 5ESS telephone switch. It took the cord that ran between adjacent cabinets and replaced TDM with VoIP. It didn’t add any features or capabilities, lower costs, or even simplify interfacing to third-party VoIP service providers. It just enabled Lucent and its carrier customers to say that there was VoIP in the switch. And their Web site listing of features and benefits accurately named the two reasons for using VoIP: “Access to capital” and “Wall Street image.”
Not Long Distance, But Any Distance
The Internet’s business model is one in which there is generally no retail charge for usage, but almost all of the retail revenue is collected out of fixed subscriber fees. Access to the network is everything; usage is simply part of the bundle. True, in some countries, ISPs routinely cap monthly usage or charge for excessive gigabytes, but in the United States, the only limit to usage is typically some unstated “excessive” value that the very heaviest users might incur. Price is a result of cost and competitive pressure. Regulators need not apply.
To an economist, this isn’t necessarily the most efficient possible scheme. End users have no incentive to keep down costs at all. So an ISP’s subscriber may choose to watch live streaming camel races from Kazakhstan, tying up costly international pipes, while file-sharing programs like BitTorrent will grab from anywhere in the world, oblivious to whether or not a local copy is available. Location, after all, is not obvious on the Internet. Edge-based content distribution networks like Akamai have to take complex steps to work around that in order to improve performance. Any distance appears to be the same as any other. Contrast this with the PSTN’s Chinese Wall rate distinction between “local” and “long distance” calls, and the complex mechanisms designed to maintain the distinction.
PSTN Intercarrier Compensation Is Frightfully Complex
Not only does the Internet’s simple retail price model differ from that of the telephone network, but its wholesale inter-provider model does too. In the domestic telephone world, intercarrier compensation is a complex mess based on a web of regulatory classifications. Local calls are handed off between carriers at one price, known as reciprocal compensation, while long-distance calls are subject to switched access charges. ISP-bound calls are subject to a third rate, while wireless calls are classified based on a different set of geographic rules than wireline calls. Interstate and intrastate rates, of course, differ. But every carrier pays the same rate per minute or per call for a given class. So a Cox Cable, with nearly half the residential phones in some markets, pays and gets the same reciprocal compensation rate in each state as a 500-phone Mom and Pop CLEC.
ISPs Meet in a Free Market
The ISP business, on the other hand, has no regulated rates. An ISP just makes the best deal it can with other ISPs. The top tier of backbone ISPs, known as Tier 1’s, exchange packets with one another at no fee — interconnection is seen as worth the same amount to each side. Others may get free peering, between each others’ own customers, but pay for transit to other ISPs. But just what it takes to get a peering or free transit deal with another ISP is not generally public; it’s a matter of negotiation. Smaller ISPs pay bigger ones for upstream service, which gives them access to everyone else. That may be priced by the size of the pipe, but also by the volume of data exchanged, typically during a busy period of the day.
What’s not there is a distance component. To some extent that’s an artifact of the Internet’s nature. Unlike the telephone network, in which everything is designed to create billable events, such as calls, which are carefully recorded, the Internet exchanges individual packets by the billion. While telephone numbers have geographic identifiers such as country code and, in the United States, area code, IP addresses are harder to pin down. So actually billing like the phone companies does would be really, really hard. It would break the Internet’s popular business model, and raise costs substantially to do so. The cost of billing would exceed any savings or new revenues taken in.
The Great QoS Controversy
One of the big differences between the two models is the concept of Quality of Service (QoS). The Internet was designed to provide only “best effort” QoS, which is a euphemism for “no particular effort.” Every packet, in theory, is treated the same. The network is connectionless, so there is literally no memory of when the last packet came along on the same flow, so there’s no monitoring of, assurance of, or charging for, the end to end data rate. If the packets get through, great; if not, then it’s up to the end systems to retransmit. This is handled in TCP/IP by the TCP’s highly-adaptive retransmission and rate-adjustment mechanisms. It works really well if everyone plays along.
Common carriers, on the other hand, have a history of selling different QoS. A phone call’s QoS provides a fixed rate (usually 64 kbps) at very low loss and very low latency. Other network technologies like Frame Relay and ATM
(Asynchronous Transfer Mode, still used in most ADSL networks) had even more options, such as allowing a Committed Information Rate (low loss up to that rate, “best effort” for additional load), and allowing the user to order a “real-time” (short queue, low latency and jitter) or “non-real-time” (longer queue, higher latency and jitter, potentially lower loss for bursty traffic) connection. ATM and Frame Relay were designed for these options. They’re not fashionable nowadays, though. The Internet’s answer to Frame Relay is called MPLS
. It’s quite similar in practice, but its vocabulary is different, and it doesn’t specify QoS the same way. But these are all QoS-enabling technologies that are available off the shelf.
Trying to shoehorn QoS onto IP is a different story. Since IP is connectionless and QoS is a property of a connection, the network has to square the circle. It’s not pretty.
But the business model is what’s really different. The phone company model is all about selling telephone calls. That’s charging some number of cents per minute for 64 kbps. If they sold higher-rate connections for a lower price, they’d be cannibalizing their bread and butter! And many QoS options require dedicating capacity.
So if an Internet user downloads a big file — say, a movie, or a Linux distro DVD — at 4 Mbps, there’s no charge, but there’s no QoS — if the network is too busy, it will slow down. Packets will be dropped and TCP will adapt. If the user’s ISP doesn’t have great connectivity to the source ISP, then the rate will just be slow. You get what you get. The Telco’s business model accommodates QoS and rate guarantees, but they come at a price, which few seem willing to pay.
Indeed that may have been the real reason why Broadband ISDN
and ATM networks never really caught on. They could offer high speed and multiple QoS options. But in so doing, they provided an opportunity for detailed usage billing, and the voice-friendly QoS options were head-on competition for existing telephone company services. Pricing ATM low would thus have cannibalized revenues, but higher prices wouldn’t have attracted users. As a result, they punted, and let the Internet cannibalize their business instead. This reminds me of a large computer company I once worked at, that didn’t want to build PCs that cannibalized their more profitable minicomputer business. They’re no longer around. The PC business model killed the older players who didn’t adapt. Intermodel competition is not unique to telecom.
A Really Complicated Non-Solution
The Internet’s business model is based on based on abundance and the bulk exchange of packets. The big telephone companies like their old business model and don’t like the Internet’s. So they’re setting out to tame it, to capture it, to replace it with something they can believe in — and bill for.
That’s the general idea behind IMS — the IP Multimedia Subsystem (News - Alert). Created for the 3G wireless business (speaking of multibillion dollar disasters), it is now being marketed to phone companies as a way to turn IP networks into collections of billable “services”. The Internet’s so-called “end to end” architecture, where the network carries packets on a “best effort” basis with no regard for content, is turned on its head. Information is still encapsulated in IP packets (gotta have that Wall Street Image!), but the network maintains an iron grip on everything. All packet flows across the network are authenticated and logged so they can be billed. QoS is imposed within IP: Capacity flow across the network is parceled out based on how the network thinks the “service” should get it.
The network can even perform “deep packet inspection” (DPI) in order to determine what’s inside a packet, how much it’s worth, how to route it, what QoS to apply, and how to bill for it. Now in the past the word for that was “wiretapping,” and could get you into trouble with the constabulary, but the telcos think they have gotten around that detail. The Internet itself is not a telecommunications network. It’s classified under American law as an “information service,” which means that the network is allowed to dig deeper than in plain “telecommunications,” which just provides the raw transport for packets. In the IMS world, the network takes over the application itself. Your own communications becomes their information. For instance, even if you don’t use the network provider’s own e-mail server, the network could charge you for every spam you fetch from your own mail service provider. If you download music, it can take a per-song fee, even if it isn’t your choice of music store. And of course it can bill for a VoIP telephone call, just for passing over its wires.
The usual justification for IMS is that it allows the network to provide superior Quality of Service (QoS) options to IP applications, such as telephony, for which the usual service is perhaps inadequate. It doesn’t do this by allowing users to request QoS, the way it worked in the telecom model; it does this by deciding what QoS it feels like giving the user. Fitting QoS onto an IP network is always difficult. Doing it the IMS way adds a Rube Goldberg-esque level of complexity. Its main beneficiaries are the big equipment vendors who can sell years of system integration service to the rubes, er, carriers who are trying to get it to work.
But why would even the Bells want to do something this screwy? The answer is obvious — IMS bends IP to fit the telephone business model, where everything is a billable event, the network operator keeps control of the applications, and it charges for the value of what’s transmitted. It doesn’t use Internet Protocol the way it was designed to be used; rather, it turns IP on its head and abuses it to do exactly what it was not designed for. Indeed, it’s more like ATM built out of IP components, as if it could cannibalize the Internet.
Adding insult to injury, IMS does this using SIP (Session Initiation Protocol (News - Alert)), a protocol designed for end-to-end (Internet model) VoIP applications. And the network’s support for public Internet applications can become marginal at best — one DPI promoter called it “hobo class,” something unbillable reached via a thin pipe, if at all. To a Bell, this is an attractive way to intercede in IP flows of its subscribers and impose the telephone business model upon them.
That’s why competition in the provision of ISP services is so important. It’s unlikely that many people would accept an IMS substitute for the Internet they’ve come to know and love. It may be “broadband” but it’s still a wrong number. If there were true intermodel competition, the telephone companies would lose badly to the Internet. So the Bells have gotten the FCC to chop the legs out from under the independent ISP industry.
Only the cable companies stand in the way. Most have shown little inkling to do something as risky and costly as IMS (though CableLabs has begun to craft an IMS-based model for cable telephony, which works perfectly well today, QoS included, using non-IMS PacketCable standards). But they tend to charge higher basic rates, and don’t serve all of the same places, so having them as the sole competitor may not be enough to kill the Bells’ plans. The Internet itself might survive this duopoly, but it’s not clear that most Americans will remain a part of it.
Most Americans now take the Internet for granted, and the rest of the world has gone even farther to adapt to its new realities. Regulatory foot-dragging and protection of incumbents has harmed the American economy and its ability to compete. But the telcos’ obsolete business model can’t last forever. The big question is how much damage will be done in the meantime.
Fred Goldstein is principal of Ionary Consulting. He advises companies on technical, regulatory and business issues related to the telecommunications and Internet industries, especially in areas where they overlap.
Don’t forget to check out TMCnet’s Whitepaper Library, which provides a selection of in-depth information on relevant topics affecting the IP Communications industry. The library offers whitepapers, case studies and other documents which are free to registered users.
Today’s Featured Whitepaper is titled Micro Dynamic Cluster: Reliability and performance for SMB’s brought to you by CommuniGate.