Presented at MIT Workshop on Internet Economics March 1995


This paper examines how the allocation of costs in the telephone distribution plant might change with the introduction of new infrastructure models—fiber optics technology and Broadband ISDN control mechanisms. We suggest that as variable costs approach zero with a B-ISDN network, a flat rate for access only may be the only practical pricing method. Pricing per bit, usage, or by differential services can only encourage the sophisticated user to manipulate future software reconfigurable systems for various types of arbitrage.

When you introduce computers into a telephone network, the network does not just change from analog to digital—with faster switching times. A digital, computerized telecommunications network becomes more like a computer: dynamic, user programmable, and less predictable. We saw all this before in a different context—with the addition of timesharing software and telecommunications channels to standalone computers, the computer changed from a batch processing, centralized environment to a distributed, more accessible networked system. We should expect changes just as radical when the network itself becomes a computer.

Former infrastructure models for both analog telephony and digital telegraphy established policies that maximized access while raising sufficient capital for future plant. With the decline of common carrier regulation, and increased competition from alternative technologies, the system is being driven towards cost-based pricing and short-term planning. However, with oncoming computer-driven ISDN, fiber-based B-ISDN, and massive capital plant recapitalization, cost-based pricing price models guarantee neither sufficient access nor sufficient capital for the future.

1. The Problem

When computers are introduced into the telephone system the network does not just change from analog to digital, with faster switching times. It becomes more like a computer: dynamic, user programmable, and less predictable. We have seen this before, in a different context. With the addition of timesharing software and telecommunication channels to standalone computers, the computer changed from a batch-processing, centralized environment, to a distributed, more accessible networked system. We should expect changes just as radical when the dial-up network itself becomes a mythical computer. Whether this network will be truly open depends upon the success of non-proprietary standardization efforts.

The implementation of an Open Network Architecture (ONA) regime [1] and Integrated Services Digital Network (ISDN) technology implies a reallocation of resources among public and private networks. This new telecommunication system is sufficiently different from the past that the transition to a digital infrastructure may be neither smooth nor equitable. Ultimately an end-to-end digital architecture will emerge, with virtual, logical connections defining communications channels.

As telephone networks become complex computer networks, user control—not carrier control—increases. With the user taking control of routing, bandwidth allocation, and administration, there will be unexpected applications of the public network, including rate arbitrage on an international scale!

The allocation of fixed costs in the local distribution network has increased in response to technological innovation, new capital and depreciation requirements, and changing regulatory regimes. The policy question is: who pays for the costs of modernization, and how is it to be implemented? Infrastructural choices must be made, and some of the economic models that were found to work well in the past, but are unsuitable now, may paradoxically soon work again.

Former infrastructure models for both analog telephony and digital telephony generated policies that maximized access while raising sufficient capital for future plant. With the demise of the old common carrier legislation, and increased competition from alternative technologies and newly built distribution plants, the current system is being driven towards cost-based pricing and short-term planning. With oncoming ISDN and fiber-based B-ISDN commitments, requiring massive plant recapitalization, cost-based pricing models guarantee neither sufficient access nor sufficient capital for the future.

No one can predict where the ISDN demand and supply curves will meet. Given the uncertain market forecasts, and huge investment in new plant, the carrier's strategy is to maximize both scale (user access ) and scope (service flexibility). This plan offers them but a mixed blessing. Not being able to predict precisely how a network will be used is an argument for building a general-purpose network, which is what an integrated digital network should be. But integration does not fit into a regulatory scheme where the network tariffs are specific to the services.

There is no way to have a service-specific tariff for a virtual, software-defined, accessed, and controlled network. And, as networks interconnect, there is no way to separate the sender from the receiver, and the ownership of the 'value-adding' portion. In the past you could locate your 'tax' on a physical part of the information plant. For example, the earliest tax on information was a license charge per page for each impression made by the printing press. In telephony, the equivalent is a metered charge based on the theory that each person's voice throughput is equivalent. Profit maximization may fit the goals of a private monopoly, but hardly exhaust those of a public service operation.

Public service obligations require that during the transition, digital/virtual and analog/specific networks must coexist and interconnect. [2] If ISDN pricing is not competitive, potential subscribers remain on the older systems. If charges are 'just right' to attract ISDN customers, we have a paradox: the faster and more programmable the network becomes, the more the subscribers compete with the carrier for the added value portion of the service. In other words, the older voice and low-speed data revenue streams 'vanish' into the interstices of the faster digital network. If the charges are fair for the older systems, ISDN never has a chance to attract traffic. You cannot have a general-purpose network with service-specific charges.

Furthermore, the dynamic allocation of network resources will become increasingly difficult to meter and expensive to track by the carrier (or regulator). With integrated digital networks, the flat-rate, or pay-in-advance subscription solution, may be the best method of pricing.

2. Models

To see how tariffing will change, it is useful to compare elements from the old and new models of telecommunications infrastructure. Whether rational or not, it is common to find that the regulation of new technologies is patterned after some older one which it appears to resemble, at least at first. [3] The old system of analog telephony was patterned on a 19th century rail transport model. Fortunately, the telephone analogy was to intercity rather than 19th century local transit. That (the streetcar) model, in most cities was no piece of cake: multiple companies, with multiple fares, and no interconnections, either locally or intercity.

This telephone wire/rail model was worked out both architecturally (trunk and branch/loop), politically (essentially one carrier and one service, except at major nodes), and for tariffing purposes (distance and time-sensitive variable costs for interregional links). It is of interest that the Federal Communications Act of 1934 was carried over almost word for word from the Interstate Commerce Act of 1888, replacing transport language with the relevant communications language, and merely adding the 1927 Federal Radio Act wording for spectrum management!

Under the common-carrier model, telephony in the United States evolved to meet the following policy objectives: universal subscription, end-to-end plant capitalization, customer price stability, and final responsibility for telecommunications infrastructure from security to R & D. The policy implications are illustrated in Table 1 below.

Though telecommunications traffic patterns and applications have changed over the decades in the United States, especially with post World War II suburbanization and the further growth of decentralized management and production,[4] this railroad transport model was, until recently, rarely challenged. Beginning in the 1960s, the telephone network underwent a fundamental transformation by incorporating digital connections and signal processing. Adoption of digital computer technology was essential for modernization and economic efficiency.

The current network is computer-controlled (SPC), provides different levels of network access, and connects sophisticated customer-owned equipment including other networks. As new services other than 'plain old telephones' abound—at least in name and in advertising vaporware—the available rate structures vary tremendously. They range all the way from lifeline flat rates to complex Centrex packages. As a system, today's 'telephone' network is very different from the original wireline-cum-transport model.

Table 1. U.S. Common-Carrier Model of Telephony
State-mandated universal service, but with business and residences splitting the cost of the fixed plant according to a Ramsey pricing scheme—differential pricing for essentially the same service. Business paid value of service, while residential prices reflected ability to pay (even if below cost).
Capital was raised by the 'dominant' carrier for the entire, end-to-end system.
Cost-engineering function averaged for the network as a whole, but the network was artificially separated into state and federal domains, which reflected the fact that calls were predominantly local.
Rates were based on plant investment. With most calls being local, unlimited flat rates for local service stabilized revenue projections.
No creamskimming. To prevent erosion of the rate base, no resale, and no private attachments were permitted.

The assumptions of the old integrated national network included: fixed overhead, fixed bandwidth, physical analog connections, and a hierarchical network architecture. The new public and private networks that are being built are increasingly fiber-based and digitally switched. The new technology brings with it a new set of assumptions, including variable bandwidth allocations with logical, instead of physical connections and a non-hierarchical network architecture with distributed processing nodes and terminals, all under shared carrier/customer control.

On the old network it was possible, originally, to separate embedded local plant from interoffice and interregional plant. That physical distinction became more artificial as more integrated equipment evolved. Still, the economics of regulation and the corporate structures of the telephone companies encouraged this artificial separation for local/long-distance, and across state boundaries, even where it was not in the public interest and probably not even in the companies' economic interest. In the future, as single-mode fiber makes possible even greater efficiency from resource sharing and intense use of all-digital networks, such separations just for accounting purposes will be almost impossible to justify rationally justify

3. Infrastructure

Changes in technology and social policy are already causing stress to preestablished costing and regulatory pricing practices. Bypass, resale, and access charges to the network exemplify this readjustment. Radical, sudden shifts in regulation, customer demands for network control, and the competitive push for new plant investment may even threaten the major carriers' viability. (See Table 2)

Table 2: Network Trends
Technology Virtual end-to-end connectivity
  • user-to-machine
  • machine-to-machine, and
  • user-to-user communication
Faster and faster computer processors
  • intelligent CPE and "smart" receivers
  • merged "packet" and virtual circuit switching
  • variable bitrate ("asynchronous transfer mode" Ð ATM)
Shared process & network control
  • between carrier and subscriber
  • line of demarcation blurred
Policy Implications

Increased international competition for service offerings

Mandatory interconnection of public/private networks


Flexible choices among technologies, carriers, rates, and services

Variable pricing and tariff arbitrage

Decline of network integrity and national sovereignty

The most efficient network designs for future highspeed, broadband, integrated digital communications should not resemble the telephone plant of the past either in architecture or functionality. If current trends continue, network usage will be different, more dynamic, and less predictable. To maximize overall system efficiency, customers will demand direct access to network resources, including operations, administration, and maintenance. Efficient computer interworking requires transparent standards. Hence it will become increasingly difficult to pinpoint whether the customer or carrier is conducting a digital transaction. At any node, the 'customer' may be another network or even a competing carrier. This has already begun: in economic terms, 'equal access' has merged carriage with the commodity being carried.

To maximize the profitability and use of the existing plant, the carrier—whoever actually owns the link or switch—will have strong incentives to expand into new areas of business ... often in competition with its own customers or other users. To maintain universal access, while at the same time expanding network provision, requires innovative rate making and novel depreciation schemes. So far, public inertia, a regulatory lag, and defaulting to yesterday's tariff formulae has artificially constrained current technological reality, making the economic transition even more painful than it need be.

4. Pricing Your Processing

Voice telephony was intended for everyone, at a cost that local residents and businesses could afford and subscribe to for life, paying one month in advance. Truly differentiated voice services under the original telephony model—as contrasted to differentiated pricing—was too difficult and too expensive to provide. Before computers, billing for any disaggregated customer base was not feasible.

Then came time-shared data processing—a special, new, and potentially 'enhanced' revenue generator. The general subscriber (and regulator) neither understood (nor was prepared to pay for) access to the universal 'computer utility.' [5]

As it happened, computing has been unregulated, and telecommunications heavily regulated. Now that the two businesses have been integrated from a technical viewpoint, regulatory attempts—Computer I, II, and III [6]—have tried (and failed) to resolve the inherent anomalies of telephones that compute and computers that communicate by physically separating infrastructural elements. It barely worked for transport, but with invisible electronic memory machines, it is simply impractical.

All digital telephone switches are general-purpose computers and therefore data processors. With the network doing de facto processing, pricing becomes much more difficult than with mere transport, as in the original telephone model (based on an analogy to physical transport).

Deregulation is easier to pull off than effective 'reregulation'! The Computer Inquiries were for the regulatory convenience of keeping data processing and telecommunications separate, at least as far as the businesses' accounting and revenue base is concerned. But it has created even more economic and political distortions: Neither regulators nor carriers have devised an acceptable plan for pricing a shared process connection between computer peripherals and the distributed database network.

5. Pricing For Capital Formation

According to classical pricing theory, in the simplest case where a single entity providing an indistinguishable basic service, price is driven to equal marginal cost. If prices are lower than marginal cost, customers tend to buy more of the service. Though volume increases, if price does not cover the cost of incremental plant, a firm (or service provider) loses money on every sale. [7] Yet, local regulators like marginal cost pricing because it tends to maximize social goals, such as universal telephone service.

Pure marginal cost pricing for an expanding utility produces a deficit which can be made up in several ways, as listed in Table 3.

Table 3. How Deficits are Covered
Through cross-subsidies (capitalization from within the system); or,
by levying special assessments on users to pay for modernization (taxation); or, (catastrophically) when all else fails
permit an obsolescent entity to go bankrupt and then nationalize it at bargain values; alternatively, if it is a bankrupt state entity.
de-nationalization, at bargain values (a political measure); or,
some other form of corporate re-organization ("voodoo" refinancing).

In the short-term, the first and last options are the least unpalatable, and a rich assortment of such schemes exist in many nations today. Telephone and telegraph services may have begun as natural monopolies, but regulation, the threat of anti-trust, and increased competition, each in its own way, have prevented too much in the way of monopoly rents. In theory, cross-subsidy, according to classical economics, would be inefficient, except perhaps as direct cash flows to the indigent user. In practice, of course, capital formation for expansion without sufficient profits would have been even more difficult without initial or subsequent cross-subsidies. Indeed, nationalized firms have had difficulty raising capital, until some financial re-organization convinces the investors to contribute more funds.

How can a new network service be extended to all those willing to pay at least marginal costs if we hypothesize that telecommunications were to have indistinguishable service characteristics? The answer, in terms of regulatory checks and balances, is in Table 1. Except for the CPE and resale provisions at the bottom (these are very important and will be discussed in the arbitrage section below) some common carrier package may return with future B-ISDN systems.

How fast are we getting there? Not too fast. So far, demand for data processing by businesses has grown at three times the rate of the demand for residential POTS. The internationalization of the market for telecommunications and information services increases business demand for network digitization and the capacity to provide faster connections, end-to-end connectivity, and especially greater access to network resources. Customer-provided computer-based terminals, and private switches, for easy connection to an all-digital network, shifts a significant portion of what would have been former carrier costs to the user.

Prospective deficits can also be filled by disaggregating services—finding some with increasing costs to be priced above marginal cost for different market segments (PBXs and key sets for businesses; POTS for residences). Monopoly privileges help enforce such disaggregation.

One basic factor in marginal cost pricing is that telecommunications, so far, has been a declining cost industry—where marginal costs drop as production is expanded. Rapid technological change in telecommunications, even without expansion, appears to reduce marginal costs even more than average costs. If marginal costs are below average costs, and price is set to marginal cost, total revenues will still be less than total costs.

Yet, despite declining marginal costs due to technological advances, the telecommunications link traditionally had not decreased in price as fast as advances in electronics had lowered computing costs and increased computer processing speeds. But a significant change is in the offing: with photonic switches and with highspeed fiber connections, telecommunications throughput may finally overtake most computer bus (input/output) speeds. For the first time, network resources may be faster than the terminal equipment can support.

This is what historians call a paradigm or 'model' change. Not only could this reverse the historical trends of computing and communications costs, but demand profiles will be greatly different than those of today. For example, new demands for fiber-based telecommunications, may come from (fiber-based) very highspeed, local-area network users. So the demand for seamless ISDN standards—and improved access to dial-up public network resources—may initially come from today's private network enthusiasts. This is something very difficult to put into predictive, linear demand models.

Altered demand profiles, no matter how potentially profitable, are a mixed blessing for carriers that have been gearing up for a different market. Despite rosy quarterly returns, a rapidly expanding and changing computer-communications plant should flag a warning that there may not be enough money for tomorrow. [8] Capital recovery for modernization may become a more critical telecommunications issue than whether basic or enhanced services should be offered, or what color and label cheaper phones should have.

6. Arbitrage

Integrated digitization of the network will not permit temporary cross-subsidization and price discrimination among customers who have the know-how and resources to bypass. The resources may be economic, technological, or political. Because control of the bitstream is shared between the carrier and the customer (and the customer may be another network), the regulatory distinctions between user and carrier become as blurred as the separation between services. With greater user control, the subscriber can balance use of the network resources versus greater investment in CPE. Reconfiguration of 'fungible' networks will be easier on ISDN systems because of their inherent fast connection times and digital access.

When further evolution toward implementation of B-ISDN, variable-rate, networks proceeds, network control may become the most important commodity being bought and sold in the telecommunications market.

Under these circumstances, how is it possible to justify different network pricing for data versus voice when these are physically indistinguishable bits on a fast link? Continuing to charge differential prices only leads to inefficient use and network arbitrage.

Arbitrage occurs when there is a discrepancy between price and cost, yielding an opportunity for a third party to profit by reselling. In some circumstances this is not quite legal, but there may be some way around it. Arbitrage is a market concept, which has not been previously applied to telecommunications; but with deregulation and virtual, end-to-end digital networks, as we will attempt to demonstrate, arbitrage is increasingly viable as long as carriers maintain a significant separation between actual costs and charges to the customer.

The new technology of integrated digital systems could result in a distinct paradox when it comes to tomorrow's carrier revenue. Though it may appear on the surface that putting all telecommunications services onto one network enhances the concept of natural monopoly (and monopoly rents for captive users), the integration of these services into just one form of carriage—an invisible, and non-distinguishable digital bitstream—has created economic pressures for cost-based pricing.

How can this revenue paradox be? ISDN promoters claim that the technology will permit the abolition of such things as private lines and other forms of competition to carrier monopoly. The answer lies in simple economic common sense: you cannot charge an educated customer more for one service when you are offering a cheaper substitute (albeit under a different name), at least not in the long run.

Some carriers currently view ISDN as a way of offering value-added data processing services—protocol conversion and information manipulation. They see this as a new way of establishing new multi-part (Ramsey) tariffs based on differential value of service. Where business uses dominate, service kinks like call forwarding are priced at higher than the marginal and average costs of their digital software. Where traffic sensitive rates apply, average costs should then replace marginally priced flat rates for private lines and local calling.

But there is a flaw: the new technology brings more than tariff and price changes, it brings incentives for usage change. With ISDN, or any all-digital switched network, subscribers have the option of only paying for pure bit transmission and providing for most enhancements with their own resources, wherever arbitrage makes it worthwhile.

The customer may be given some opportunity to whipsaw the carrier (for a change). There is little in enhanced services that a carrier can offer that a sophisticated user cannot get from well-programmed customer-premise equipment (CPE). Computer-based CPE technology, therefore, will drive carrier pricing close to marginal costs. With broadband interfaces (following open architecture principles) part of the frame overhead must be customer accessible. If carriers restrict use by monitoring content and terminal type, this would likely drive the largest users to build their own networks which will offer flexible, dynamic, frame-structured wideband services.

7. Distributing Bypass

A basic ISDN principle is that an ISDN voice call between two points will not cost any more than a voice call over the conventional public switched telephone network (PSTN) between the same two points. Attempts to charge more for a data call than a voice call on an integrated digital network will be impossible to enforce, since the bitstream will be indistinguishable by the carrier or regulator.

Equal charging will make it possible to seize a 64 kilobit digital line, at the bargain-basement voice rates. If the carrier needs to know, you tell them (digitally, of course) that you are 'talking' to your customers, but really your firm subdivides the circuit into two, four, six or more voice/data circuits, using sophisticated voice and data compression equipment already on the shelf. The carrier will be charging for only one voice circuit and you can hang onto it all day, creating a virtual, multi-circuit, tie-line.

The rate elements that were used to price services on the local loop, will no longer suffice. Rate elements based on: 1) call frequency, 2) duration, 3) distance, and 4) time of day assume costs of switching and trunking plant (interoffice and tandem) are variable. Features and functions like call-forwarding, speed-dialing and call-waiting, added a new variable element based on SPC memory and processing increments. But ISDN and B-ISDN technologies are not straight-line extrapolations of today's Stored Program Control (SPC) circuit switching, especially in the packet mode (for voice and data).

Memory tables are inherent in any case for D-channel and common-channel ISDN signaling. Adding features is more a matter of toggling bit positions in already existing memory space. Memory space enhancement, therefore, becomes virtual and depends on signaling under user control, especially where disparate terminals must negotiate before a 'call' actually begins. [9] In B-ISDN, as we already mentioned, user signaling is even more powerful, including routing and temporal feedback information for possible OA & M purposes.

8. Broadband Service

Yesterday's discount dealers have been preparing for Picturephone and tomorrow's customers have been buying into a cheaper-than-voice-telephony electronic mail service. Occasionally technology push and demand pull are synchronized

Who needs broadband ISDN (B-ISDN)? Predicting future demand for services that do not exist borders on witchcraft. But the curious thing about B-ISDN is that we may be paying for it in any case. The fiber is being installed for the trunk and local distribution plant today. How—and whether—the digital bitstream on these fibers is switched, and what strategy may be pursued to reach the end user, are a matter of current debate. Nevertheless, broadband capability is part of the natural evolution of the telephone network; only the details remain to be worked out.

But broadband does not mean bandwidth-wasteful services, it only means broadband-capable speeds. End-to-end, highspeed digital communication may generate totally different demand profiles. In the network of the future, most of today's favorite services may not require continuous channel occupancy. For example, movies can be encoded to be sent over medium-bandwidth lines in compressed time (not in real-time on the network). Packetized voice needs hardly any bandwidth at all. [10]

Virtual networking entails network transparency for optical broadband interfaces. [11] B-ISDN is an order of magnitude more powerful, with potentially a steeper decline in transmission and switching marginal costs. Single-mode fiber can upgraded almost without limit by adding more interface electronics; it is not necessary to string new links. This fact, along with the possibility of altogether new architectures for superfast switching has led to an effort to proposals for B-ISDN to leapfrog narrowband ISDN technology. [12]

Broadband 'fast-packet' networks does not imply that services will demand continuous holding times. Most services tend to be intermittent, and fit well in the technologies proposed for switching fiber optic networks. Television, as well as, can be asynchronously transmitted. This is a very important technological concept which underpins B-ISDN pricing schemes. Despite data which indicates that HDTV requires hundreds of Megabits per second, this does not mean it needs that bandwidth continuously! [13]

The development of 'fast-packet' switching, and especially single-mode fiber makes B-ISDN an altogether different beast than narrowband ISDN (N-ISDN). N-ISDN is a end-to-end digital solution offering a basic rate of 160 Kbps, channeled into 2 'B' (64 Kbps) and one 'D' (16 Kbps) streams (conventionally, 2B+D), the D channel ostensibly for signaling. [14] A primary rate of 1.5 or 2 megabits per second is also specified. These rates were originally intended for existing twisted copper-pair plant, though the primary rate uses mostly coaxial cable or fiber today. And moreover, it may be that much of the copper plant has to be rebuilt because the old wires are not conditioned to carry the 2B+D digital rate below acceptable levels of electromagnetic interference. [15] Should this be the case, rebuilding with fiber may be the more economically expedient for the long run; broadband ISDN, therefore, may leapfrog N-ISDN in many areas.

B-ISDN standards apply sophisticated relational database software across a high-speed digital link for both transmission and routing. Hence it subsumes the narrowband 2B+D rate. Switching under this schema gains enormous power, but can be under the control of the subscriber, instead of just the carrier. The packets and signaling system are combined so as to be essentially self-routing.

The key to network design is the use of enormous bandwidths for overhead made possible with fiber. Current CCITT standards call for an H4 rate at about 150 megabits per second. The 'payload' is 90 times larger than the primary rates, or 1000 times the basic rate for a local loop. Indeed, overhead alone is about ten times larger than the entire payload capacity of primary circuits.

The new interface arrangement is in the form of a matrix, replete with pointer cells. It is these pointers which allocate the system resources. This matrix, called a 'frame,' can accommodate various headers, and sub-headers in an 'envelope,' so that these very high-speed packets or envelopes of frames can be self-routing (independent of nodal 'hierarchies'). Variable rates could be transmitted under an 'Asynchronous Transfer Mode' (ATM). [16]

Envelopes permitting variable bandwidths give users powerful network control options without reducing the carriers management and operations control. The new software-controlled options could be designed to permit carriers and customers to share control of a 'virtual' network, without mandatory co-location. (This is what the FCC's Open Network Architecture proceedings should be all about.)

The H4 channel would accommodate two digitized, compressed, but continuous high-definition TV (HDTV) signals, or 4500 uncompressed, simultaneous telephone calls. But, as we noted, sending such data in real-time is an insufficient use of ATM framing. The beauty of the frame matrix structure is its elegance in handling multiple services and bandwidths in very short time intervals (125 microseconds per frame), facilitating video in delayed time, or data in compressed time, or packetized voice. Again, we emphasize that B-ISDN does not necessarily mean wideband, continuous occupancy of a channel.

These drastic ratio differences between narrowband ISDN (and analog telephony) and broadband ISDN are bound to affect tariffing theories and increase the range of user-defined services. Moreover, as long as carriers have to put in single-mode fiber for the distribution network, and are willing to adopt B-ISDN frame interfaces, it makes little difference in cost whether the packets are running at 151 Mbps or 1 gigabit per second!

9. The Universal Flat Rate

Figure 1 (below) suggests that the true variable costs will approximate a step function based on incremental costs for bandwidth, dependent on the laser access to the fiber (i.e., the remote electronics). Access to the network via the 'local loop' would be based on the maximum instantaneous bandwidth a subscriber requires. This might be the simplest, effective way to tariff broadband transactions. (In addition to access, a charge may be levied for degrees of non-blocking or priority service.) While the average user of POTS may not need to access all of the Library of Congress online via B-ISDN, video-on-demand could well justify a flat rate for fiber access to the home. And it is likely that even narrowband ISDN will find that access is a more productive policy for capital expansion than policing bit-per-second usage.

Figure 1: Future Cost vs. Speed for a Future B-ISDN PlantFigure 1: Future Cost vs. Speed for a Future B-ISDN Plant

Until 1980, 80-90% of the costs of the Bell System plant were joint for local, short-haul and long-haul services. [17] On a distributed computer network like the Arpanet, or even Bitnet (where ownership of each link is somewhat mysterious!), the users mainly pay for access; usage costs vary depending on their CPE architecture. Like these relatively slow-speed, customer/carrier shared packet nets, future transparent 'fast-packet' networks, based on SONET asynchronous frame transfers (working their way at Gigabit speeds through a mesh instead of an hierarchical network) will find variable switching costs too small to measure or detect. ATM architectures make pricing the whole of the network relevant to any local tariff scheme. Local will be inseparable from medium-, if not long-distance in B-ISDN. Artificial costing separations, as in today's state/Federal network, will only encourage bypass.

As the separation of basic service elements increases in complexity with end-to-end digitization, the only feasible solution from the aggregate user perspective becomes a universal flat rate for access. A flat rate is even more attractive under B-ISDN. With the vast inherent capacity of optical fiber, fixed costs of the distribution plant overwhelm any variable costs. With a flat-rate solution, the sum of customer access costs then would be equal to the total revenue requirement for each carrier. This might solve the arbitrage problem and the carrier revenue paradox, to boot.

10. Paradigms

Three systems or network models were reviewed here: 1) 'carrier-based' telephone, telegraph and railroad; 2) transitional service-dedicated analog/digital; and, 3) integrated, fiber-based, all-digital and switched broadband. During the current transition period (a hybrid mix of analog/digital systems of non-universal character), the combined public and private global networks exhibit a 'duplicative functionality.' This situation permits clever users to manipulate these systems to their own advantage, thereby bypassing the common-carrier pricing rationale.

In a distributed data processing environment—which is what ISDN is all about—the old revenue requirement game may no longer work. Conventional economics is leading to some bizarre rates and service offerings. The underlying digital process subverts the previous rules, because computer network architecture does not respect artificial physical boundaries which previously regulated access and interconnection.

New tariff policies will ultimately be worked out for ISDN access. When old plant has to be amortized at the same time the new system is capitalized, painful reallocation of resources and privilege may be necessary in order to avoid future revenue losses.

An integrated services digital network brings systemic change as multiple and fungible services and resources can be reallocated in real-time. If transparent networking capabilities are realized (this requires full adherence to international standardization efforts for interoperability between different vendors' CPE and public ISDNs) carriers will lose the ability to differentiate between services. Cross-subsidization policies will then be difficult to continue.

Author Information

Richard Jay Solomon ( is affiliated with the MIT Research Program on Communications Policy, Center for Technology, Policy, and Industrial Development, Massachusetts Institute of Technology. This paper does not represent the views of MIT or RPCP.


1. FCC Computer Inquiry III (ONA), June 1986 (see Ref. 6, infra). The European Community's equivalent is the 'Open Network Provision.' See COM(88) 240, EC, Brussels, and [European] Council Recommendation 86/659/EEC, 22 Dec. 1986.return to text

2. FCC, id.return to text

3. Pool, Ithiel, Technologies of Freedom, Harvard University Press, 1983.return to text

4. The U.S. has been more decentralized than most industrialized nations ever since the later decades of the 19th Century—a socio-economic national preference aided (or abetted) by the rapid spread of first steam and then electric interurban railways into the countryside (the latter the functional forerunners of the U.S. highway system built in the 1920s); this transport revolution took place simultaneously with the growth and penetration of the telephone. This decentralized infrastructure may help explain the tolerance for the U.S. decentralized and fragmented common carrier regulatory system. Only prewar Germany approached the U.S. level of industrial decentralization; but communications in Germany, like its railways, has always remained centralized under the national government.return to text

5. Loretta Anania, 'Network Planning in the Information Society,' Chapter 1 of unpublished Ph.D. dissertation, MIT, 1988. While so far data constitutes 6% of the traffic, it has grown at a rate of 30% per year recently. POTS has grown only about 10% per year.return to text

6. Computer I: 17 FCC 2d 587 (1969) and Docket No. 16979 (FCC 66-1004) passim.; Computer II: 77 FCC 2d 384 (1980) and Docket No. 20828 (FCC-76-745, July 29, 1976); and Computer III (Open Network Architecture): Docket No. 85-229 (104 FCC2d 958, May 1986)return to text

7. Anania & Solomon, "Capital Formation and Broadband Planning: Can We Get There From Here?," Telecommunications, Nov. 1987; "Integrated Digital Systems A Threat to Carrier Revenue Requirements?," Telecommunications, Feb. 1987; I. Pool, M. Sirbu, and R. Solomon, "Tariff Policy and Capital Formation in Telecommunications," Center for Policy Alternatives, M.I.T., published in Evoluzione delle Telecommunicazioni Negli Anni 80, Tecnetra, Rome, Italy, 1981.return to text

8. See Anania & Solomon, "Capital Formation...", op. cit., for an explanation of the puzzling phenomena that telecom is both expanding and undercapitalized. See also R. J. Solomon, "Changing the Nature of Telecommunications Networks," Intermedia, May 1986, pp. 30-33.return to text

9. See CCITT "Red Book" (1984), Q.723 (SS7 Formats and Codes); I.311 sect. 9.2 (ISDN Numbering Plan, sub-addressing); and controversy on user-to-user (UTU) signaling using the sub-address field, and data transfer via 'switch-through' before call charging begins in CCITT Study Group XVIII Document R-4-E, "Report of sub-group on numbering and addressing," July 1985, sect. 7.6; CCITT Working Party III/6 on "Charging and Accounting for the ISDN and common channel signaling network," TD Nos. 609-E, 616-E, and 619-E through 625-E, June 1986, Kobe, Japan; and updates in SG XVIII Rep. R49 (A-C), Seoul, Feb. 1988 ("Network Aspects"), esp. sect. I.333, "Terminal Identification in ISDN," pp. 17ff.return to text

10. In the bit-metered, digital world, female voices can be compressed more than male, English more than German, and other inequities may emerge, especially with digital applications of 'Time Assigned Speech Interpolation' algorithms (D-TASI).return to text

11. R. J. Solomon, "Vanishing Intellectual Boundaries: Virtual Networking and the Loss of Sovereignty and Control," Annals of the American Academy of Political and Social Science, vol. 495, January 1988. [doi: 10.1177/0002716288495001004]return to text

12. For a more complete description of the issues in B-ISDN and a technical bibliography, see R. J. Solomon, "Open Network Architectures and Broadband ISDN: The Joker in the Regulatory Deck," in ICCC-ISDN'87 - Evolving to ISDN in North America.; this paper contains a detailed technical bibliography. The October 1987 issue of IEEE Selected Issues in Telecommunications is devoted to broadband switching concepts. A tutorial is in Steven Minzer, "Preliminary Special Report on Broadband ISDN Access," Bellcore SR-TSY-000857, Issue 1, Dec. 1987. See also refs. 9, 15, and 16 herein. For future evolution, see CCITT SG XI, Ad-hoc Group on New Questions, "Proposed Study Questions for the Next Study Period," TD 154 and TD142, Geneva, 16-27 May 1988.return to text

13. Sophisticated motion detection algorithms permit video information to be transmitted only when sections of an image change. Raw information can be sent in "compressed time," that is faster than it would be ultimately viewed, to recreate a full movie. See Anania & Solomon, "The Ghost in the Machine—A Natural Monopoly in Broadband Timesharing?" Telecommunications and Telecommunications International Edition, October, 1987. For a discussion of the current state of video encoding see Andrew Lippman, Vol. II Tutorials, "Video Technology: High-Definition Television," NCGA '88 Conference Proceedings, National Computer Graphics Assn. 1988return to text

14. The D channel is perfectly capable of transmitting a customer payload since the 16 Kbps rate is overkill for signaling—even during call setup and teardown. Note, in the primary rate, 64 Kbps is reserved for the D channel.return to text

15. Swiss PTT, "EMC and COMSEC problems on ISDN basic access installations (rec. I.430)," CCITT SG XVII, D.1740/XVIII, Geneva, 6-17 June 1988.return to text

16. See CCITT, Study Group XVIII, "Report of the Working Party XVIII/7 (Transmission Aspects of Digital Networks)" COM XVIII-R 44 (A, B, & C), August 1987 (Hamburg Report) and the CCITT "Red Book," I-series recommendations (1984) for more detailed information on ISDN specifications. Current material on B-ISDN proposals and standards is in Minzer, op. cit., and in CCITT reports from SG XVIII, R55 (A, B, & C), Seoul, Feb 1988 ("Broadband Task Group"). These are being codified for the 1988 Blue Book; drafts are available in the CCITT reports from the July 1988 SG XVIII meeting in Geneva. Also see ref. 12 infra.return to text

17. Oettinger, A. et. al., "Players Stakes and Politics...", Harvard Program on Information Resources Policy, Harvard University, 1981.return to text