Unleashing Ethiopia's Potential: The Technological Reasons for Open and Competitive Cybercommunications

by Nemo Semret

1st ESS Conference on Ethiopian Telecommunications in the Information Age, Washington, DC, July 2nd, 1996

Regulatory reform is the most crucial challenge facing us at the dawn of the information age in Africa. This paper will make the case for deregulation and private enterprise as a must for cybercommunications to develop in Ethiopia. While the case can be made from a wide range of perspectives (socio-economic, historical, etc.), the approach here s bottom-up, with emphasis on the technological issues, and how the necessity for an open competitive environment arises from technical considerations inherent to modern communication networks.

In section 2, we analyze the fundamental paradigm shift between traditional telecommunications (with telephony as the representative example), and modern cybercommunications (of which the Internet is the prime example), and show how while monopolies may have been somewhat justifiable in the former, they are not necessary in the latter. In section 3, we go on to show how an open competitive environment - at the bearer, access provider and service levels -- is crucial to the development of a digital information infrastructure, with particular emphasis on the issues specific to Ethiopia. Section 3.3 considers some future networking technology issues, and briefly extends our case from the medium term to the longer term. We conclude in section 5 with some broad views on the ideal role of government in the tele/cyber-communications sector of the future.

MONOPOLIES ARE NO LONGER NECESSARY OR NATURAL

Traditional telecommunications as "natural monopolies"

The early days of the telephone were far from the monopolistic situation we have seen all over the world through most of this century. The field was rich with competing enterprises, and even competing technologies. However, as soon as the pioneering days were over, when large practical networks had to be built, telecommunications revealed itself as a field for "natural monopoly".

The fundamental reason for this was the analog nature of the signal being transported. Essentially, the information is represented by a voltage between two wires, and the precise shape of that voltage as a function of time has to be carried faithfully and delicately from one end to another. While anything could go on very small networks, this had two fundamental consequences on the design of large networks.

1. The signal being transported was "passive", i.e. it contained no information other than the content itself, e.g. the sound of your voice. For example, the signal did not "know" its destination. This means that before the signal was carried, a precise route on the network had to be established from end to end. Thus, in order for large networks to function, the only option was to have a very rigid and hierarchical topology. A signal from A to B always went on the same predetermined path, for example: home, central-office, regional switch, central office, home. For new users to join the network, they had to be added to the network in a centralized manner and assigned a phone number which precisely indicated through which sequence of switches they could be reached if and when they were called.
2. Because they had to be very carefully guarded against interference with each other, it was difficult to have many signals share the same physical means of transport (wires). Once a network (or a portion of a network) was built, the maximum number of calls it could support was fixed. The only way to support more calls would be to install additional physical resources, or at least make important changes at all the switching points (e.g. increase the multiplexing). Furthermore, the network bandwidth needed to complete a telephone call (a "circuit") had to be strictly reserved from end to end before the actual communication could begin, and monopolized during the whole duration of the call.

For those two reasons (fixed routing and resource reservation), a telephone network had to be controlled in a centralized fashion, and central control implied central ownership. A would-be independent telephone service provider would have to fit into a precise place on the rigid hierarchical topology; this meant that, to have their calls routed to domains other than its own customers, they were permanently tied to the structure at the next level above them. This gave a strong incentive for the industry to be vertically integrated.

At the same time, because telephone networks were built from scratch, and each individual user had to be wired to a central office, the initial cost of entry into the business was prohibitive. These dominant economies of scale led the industry to became horizontally integrated as well.

"... weren't really suited for a competitive marketplace. The resources they needed to bring their products or services to market were simply too scarce --or so expensive to provide that the pool of potential customers could not support more than one supplier." [Notebaert]

The advent of digital technology changed the underlying reasons, but because of the size of the existing infrastructure, and the need for backward compatibility, even after trunk lines were converted to optical fibers transporting digital signals, and the central office switches became digital one after another, the centralized monopolistic structures remain, and introducing a competitive market in these areas remains difficult. It is true that technological improvements in telephony can help liberate from constraints inherited from previous generations (e.g. cellular phones in the local loop, or microwave transmission for long distance):

"advances in technology have spurred the development of competition by removing entry barriers in markets that, up to that point, had been served by monopolies" [Notebaert]

Still, in the "plain old telephone system" (POTS) competitive markets are justifiable mainly on economic and social grounds (better, cheaper service, and more choice for users), and while technological progress may be of an "enabler", the structure of the system itself is not a compelling reason for competition.

Cybercommunications as "natural free markets"

In the new cybercommunications, as represented by the Internet, the structural tendencies induced by the technology are virtually the opposite of those in the traditional telephone business. Here, the technology is in fact a compelling reason for deregulation, and free enterprise, independently of the economic and social reasons.

First, the information is, and has been from the beginning, digital. This means it can be broken up into "packets" which are sent one after the other. The packets can be thought of small envelopes, each carrying a piece of information. The receiver can then open all the envelopes, and put the information back together again. For our purposes, the two key differences from the previous case are:

1. Each packet is labeled with the sender's and the intended receiver's addresses (this is in what is called the "header" of the packet). This means that one does not need to figure out the end to end route before sending the data. Packets can be simply pushed into the network, and the intermediate points can push them along by looking at the destination address in the header. On the internet, two consecutive packets from A to B (even two packets which of the same data transfer) can even travel by different routes.
2. The packets need not have strictly reserved bandwidth to travel on the network. Unlike the analog signals, at any intermediate point during their trip, they can be held temporarily without being lost, while they wait for other packets in front of them to move on and free bandwidth. Also, they do not need to be as carefully protected against interference. If they are corrupted, either they can be restored to perfect condition through digital error correction methods, or the sender's machine can simply be told to resend an identical copy of the packet. Generally, the user will not even notice it happened.

Because routing is not necessarily fixed and hierarchical, and because resources do not have to be reserved ahead of time from end to end, the Internet is entirely decentralized. Indeed, this was one of the principal objectives in the design of the ARPAnet, the grandfather of the Internet [see Bertsekas and Gallager]. There is no central control. Each node in the network makes its own decision on whether and where to send a packet it receives. Additional control for a specific stream of data may only be done at the end user's machines.

Essentially, the only requirement for a machine to be part of the Internet (to be a host) is to understand the packet header format, to have an address, and to be connected to at least one other machine which is on the Internet. Packets reach their destinations because all hosts continuously tell neighboring hosts: "I'm here, this is my name (address); if there are any packets for me, send them here". Unlike the POTS, a network node does need to fit into a specific place in a hierarchical structure. In principle, one could disconnect an internet host, carry it to the other end of the world and connect it to another node on the Internet and within minutes, it will have alerted its new neighbors, and it could resume work exactly as before with the same address, without even missing a single E-mail that was sent in that time.

This has immediate consequences for the structure of the industry. Monopolies do not naturally develop, because it is relatively easy for a new service provider to join the network, or an old one to change inter-connections.

Perhaps most important, and least known, is the fact that the Internet is not a physical network. It is simply: a) a set of explicit agreements (or protocols) on routing, packet format and addresses which anyone who wants to join has to follow; and b) an implicit agreement that all nodes will forward the others' packets in the way they think is best to reach the destination at that time. Most of the physical structure consists of lines leased (usually from telephone companies) to interconnect the hosts.

Thus, the barriers to entry are not as prohibitive as that of, say, building a telephone central office and laying wire to thousands of homes and offices. The estimated total sales of Internet-specific hardware in 1995 were only $50 million [Anderson], far less than $2 per user per year (of course this is the figure for the network per se, not including end-users equipment such as PCs and modems). This tiny sum, even though the Internet was growing at a tremendous rate. Thus, the economies of scale are not dominant, and the Internet won't naturally become horizontally integrated, or dominated by monopolies.

Back to top

NECESSITY OF ENTERPRISE-BASED DEVELOPMENT OF CYBERCOMMUNICATIONS

In the previous section, we showed that monopolies (whether private or state owned) are neither natural nor necessary in cybercommunications. Here we go a step further, and make the case that for cybercommunications to develop in a country such as Ethiopia, an private-enterprise based market with a blend of cooperation and competition is necessary, even crucial.

That the dawn of the information age offers a unique (and rapidly closing) window of opportunity for a developing country to emerge from poverty and be successful in the global economy is well known, and we will needs not stress it further here. Also, that connecting to the Internet, currently the only functional global cybercommunications network, is the key first step in this process is taken as a given. What is not at all clear to anyone at this stage is what is the best way for Ethiopia to achieve connectivity. It is useful to separate the issue into two levels, following a standard model in networking. The bearer level, which concerns the network itself, i.e. the business of transporting information from one place to another; and the service level, the wide largely uncharted questions of what kinds of information will travel over the networks, from who to whom, and for what.

Bearer Level

Ethiopia at one time was a pioneer in communications. The first long distance line, Addis Ababa to Harar became operational in 1894, less than a decade after the first long distance line in the US [Tsigie and Feyissa]. In mid-1996, almost 3 decades after ARPAnet became operational, and at least five years after the first Internet node appeared in Africa, Ethiopia is one of the last remaining countries not connected to the Internet. Achieving connectivity is of paramount importance, and will probably occur in late 1996. However, the lag is already large, and connectivity must begin and keep increasing fast.

A wide array of very different alternatives for physical layer connectivity present themselves. The most immediate one appears to be using the infrastructure currently in place for POTS -- mainly by direct satellite links to 17 countries in Europe, the Middle East and North America [Tsigie and Feyissa]. But, already, other means abound for connections within the country and to the outside. VSAT, or small aperture satellite for connection from remote areas with insufficient or non-existent phone service; packet radio; even cellular modem access may soon become feasible for cities, as the cost of installing and operating base stations is dropping faster than that of traditional wiring to the home/office.[Davies]

In the medium term, the fastest connection to the outside may be through the Africa ONE submarine fiber optical network encircling the continent planned by AT&T, which will be ready for service by 1999, and will have trunk speeds of up to 40Gb/s [Marra and Schesser]. Another, perhaps more exciting, possibility comes from ventures such as Teledesic which plans to launch 840 low orbit satellites to cover the globe, and offer very high speed, low cost Internet access to even the remotest areas by 2002 [Kupfer].

For Ethiopia, given the very low degree of existing wiring (only 130,000 phones in 1993 [Tsigie and Feyissa]), and the fact that the vastness of the country makes extensive physical wiring unlikely in the near future (or ever), the low orbit satellites are an attractive option. Going through low orbit satellites would have low enough propagation delays (the time for the signal to travel over the link) for most applications, unlike traditional communications satellites (which are geostationary, and thus in very high orbits and so cause delays too long for inter-active applications).

Given the wide range of techniques and the uncertainty surrounding most of them (either fundamental uncertainties due to the untested nature of the technologies, or simply uncertainties about applicability to the specific context of Ethiopia), ultimately, the market would be a better arbiter than a single state run body as to what is the best combination to provide bearer level connectivity. Indeed, a number of the alternatives are bound to fail, and a number of them could successfully coexist. Can the state, which by nature is -- and must be -- risk averse, take the risks required to find the right combination? By retaining exclusive license to operate telecommunications for itself, a government could end up losing a lot of public money, unaffordable for a country such as Ethiopia, where basic needs such as food, shelter and health are still far from completely satisfied. At best, the government would choose the least risky alternative, and end up serving a limited user base in sub-optimal ways.

By exercising monopoly rights, the state would prevent creative small-scale solutions, and thus leave large sectors of the population unserved. In a nation where the general level of development is very low, the skills to use and demand for cybercommunications would best grow in diverse, and small "niche markets". Realistically, there might only be a few hundred customers for a particular type of connectivity (e.g. agricultural experts requiring mobile satellite connections in remote parts) and they would be best served (probably only served) by independent small-scale, specialized providers.

Unlike POTS, where there were originally much fewer and less diverse potential solutions to basic connectivity, for the Internet, the choices are many from the ground up, and due to the flexible structure of the network and the rapid rate of innovation, the range of choices will only continue to be wider and more confusing. Private enterprise is, by definition, willing to take the risks, and can start on small enough scales to proceed by trial and error. And because economies of scale are not dominant, they can do so even without prohibitive capital investments.

Perhaps the only strong argument in favor of monopolies is the need for universal service, and that this can only be achieved through the cross subsidies which are possible under government mandated monopolies. But, in cybercommunications for a third world nation, not only must the suitable skills and demands have been nurtured and developed in specialized pockets where the need is already felt, but also the fixed cost being relatively low, community-oriented initiatives can start and thrive even in a deregulated environment without cross-subsidies (as evidenced by the continued survival of "FreeNet" public access providers in the increasingly commercial Internet of the US).

Moreover, even in the POTS business, where a monopoly may be technically (though not economically) more justified, due to the inertia of state bureaucracy, progress toward universal service is practically non-existent. In the case of Ethiopia, the waiting list for telephone lines is growing ever faster: in 1993 it had already reached a level where, at planned rates of capacity expansion, it would "require about seven years just to clear the existing backlog of registered telephone demand" [Tsigie and Feyissa]. That was only for customers already waiting to be connected, not counting the new demand coming in during those seven years!

Service level

There are even fewer commonly accepted predictions as to what the services provided on cybercommunications networks will evolve into than there are about the bearer level technologies. The obvious example is the World Wide Web, which appeared suddenly in 1993, and by the end of 1995, had grown to eclipse everything else on the Internet (except perhaps e-mail) to the point where in everyday speech "Internet" has become synonymous with "World Wide Web". No one had foreseen this now dominant service even six months before it appeared. Another example is in Ethiopia itself, that of PADIS, the only e-mail provider. The service has been relying on a combination of FidoNet software (public domain, i.e. free software), and dialing-up to a Fido-to-Internet gateway in Europe via regular telephone lines. By creatively combining a variety of different things, PADIS (and others in Africa) are able to provide a very valuable pioneering and robust service at low costs. Yet this solution would most certainly never have been tried if one set out to develop a nationwide grand plan for an e-mail system. It could only have evolved, as it did, through small scale, hands-on efforts.

At the service level, even the strongest proponent of state control would concede that no monopoly, whether state owned or private, can ever anticipate the needs or be innovative enough to fully satisfy them [Davies]. In fact, once the bearer and access provider levels are in place, it is literally impossible to restrict the development and deployment of application services by third parties.

The service level is where the information infrastructure will be most beneficial in the long run. In addition to the oft cited developing world applications for health, agriculture, etc., one key contribution might be in spurring software development for specifically Ethiopian needs, such as Ethiopic document preparation tools (word processors, editors, etc.). Indeed one of the great obstacles in this domain currently is the pool of users being too small and too dispersed to be reachable efficiently by the "normal" application software distribution and user-support mechanisms. Thus the users end up acquiring the software through informal means, like copying from each other, which results in a disincentive for application developers.

In a networked world, since computer users will be the same group of people as the network users, they can get the product easily, cheaply and legally over the network themselves, as well as have access to the support and documentation which make legal acquisition attractive. This in turn will provide the incentives for more software development, even for the very small niche market which Ethiopia would necessarily be. With adequate skills being developed in the educational system, this could possibly lead to the development of an export industry, with the networks providing the means of export as well. This may seem like an overly rosy scenario, but it would not be the first time a highly technological industry has had Ethiopian participants being competitive internationally: the example of Ethiopian Airlines shows that given the right conditions, the required skills can develop and flourish within one lifetime.

Beyond the Internet: the next generations

One thing that is clearly emerging from research into the future networks is the need for open interfaces. That is the software boundaries between the different levels, bearer and access provider, access provider and application services, should be clearly defined and publicly known, so that anyone can build something on top of what others have done on the level below. Perhaps the best analogy for this is the personal computer world: microprocessors made by Intel fit into a motherboard originally standardized by IBM, on top of which an operating system by Microsoft provides a platform for applications software from thousands of different sources. At each of those levels, the interfaces were made public so that different companies are able to offer products, and consumers are not trapped in a system that is proprietary to one company from top to bottom. This openness is what fueled the rapid growth of the PC industry, with costs driven down by competition in price as well as technological innovation.

The NRenaissance Committee appointed by the US National Science Foundation to study the future National Information Infrastructure (NII) concluded that what was needed for the future was:

"... a data network with open and evolve-able interfaces. Such a network should be capable of carrying information services of all kinds, from suppliers of all kinds, to customers of all kinds, across network providers of all kinds, in a seamless accessible fashion."[NRenaissance]

In long run, this same vision should guide information policy in Ethiopia. No country exists in a vacuum -- or if it did, why bother with communications networks? -- and the future is towards openness and deregulation, and competition in communications. Not purely for economic reasons, but mainly for a technological reason: the need to have flexible adaptable networks that will grow and evolve to meet unforeseeable needs and opportunities.

"the ability to evolve is a key property of the NII envisioned by the committee."[NRenaissance]

The technical challenge for the coming generations of networks is to recapture the "Quality of Service Guarantees" (QoS) that are available in circuit switched networks. For example, in the phone service, there is no variation within a call of the sound quality, and there are no unexpected delays because exact amount of resources (bandwidth) is reserved exclusive for that call. Thus, in principle a call always has a guaranteed quality of service. When there are insufficient resources, the call is simply rejected (you get a busy signal - which does not occur only when the number you are calling is busy, but also, if all the circuits at some intermediate link are already taken).

The drawback is that the network is rigid, expensive to build and control (centrally), and also wasteful of resources (during a phone call, you use the same amount of bandwidth when you are silent as when you are talking). On the other hand, packet switched networks like the Internet provide tremendous flexibility, openness, and maximize efficiency in the use of resources. This comes at the expense of not being able to provide guarantees of exactly when a packet will arrive at its destination. But such quality of service guarantees will be necessary for real-time video, voice and other time delay sensitive services that future multimedia networks will aim to deliver. The challenge then is to try to get the best of worlds: a packet switched network with quality of service guarantees. This is essentially what the proposed Asynchronous Transfer Mode (ATM) standards aim for.

While most of the issues of resource allocation for packet switched networks with QoS remain open research problems, a promising avenue of research is that of modeling a network as an economy, a decentralized system of independent "agents" which compete for resources. This research approach is not about the economics of operating networks, but rather it is borrowing tools and concepts from economic theory, particularly Game Theory, to develop a way of studying and designing networks which leads to distributed/decentralized intelligence and control of the network, which makes them scaleable to large and diverse networks [Semret].

"The ability of trade and price mechanisms to combine local decisions by diverse parties into globally effective behavior suggests their value for organizing computation in large systems." [Miller and Drexler]

Thus, when market concepts are applied to engineering networks from the technical standpoint, it is only natural that the development and operation of such systems be done in a decentralized and competitive manner, driven by free enterprise, experimentation, openness and innovation.

Back to top

CONCLUSIONS

A caveat is perhaps in order. Free and private enterprise does not preclude cooperation. While competition is a healthy engine of innovation, the basic nature of networks, namely interconnection, obviously requires a substantial degree of cooperation, particularly in having open interfaces and inter-operability (a customer of one entity must be able to communicate to a customer of another).

This paper has argued that in the information age, monopolies are not only no longer necessary, but also would be harmful to the development of cybercommunications in Ethiopia. The main reason for this, we claim, are not economic (though there may be strong arguments on those grounds), but technological. The very nature of the information technologies virtually requires an environment of free enterprise and dynamic innovation.

The appropriate role for the state is to continue investing at ever higher rates in the physical national infrastructure, but allow a wide range of bearers and access providers to operate, whether leasing resources from the state built physical network or building their own alternative physical means of transport. State owned entities could also enter the bearer market as competitors, but these entities would best be clearly separated from those building and owning the physical layers.

Another crucial government role is as a regulator, ensuring the fair and efficient functioning of the market. This need is best illustrated by the following fictitious example: XYZ, an Ethiopian start-up Internet access provider cannot afford it's own link to the global network. Say the best option would be to lease bandwidth on AT&T's under-water fiber encircling Africa. But if AT&T also owns a competing access provider in Ethiopia, they could charge XYZ much more than their own subsidiary, drive them out of business, and thus end up controlling the market.

At the same time, protectionist approaches like blocking foreign ownership altogether are not constructive since that would only distort the market, and encourage corrupt hidden dealings. So we would need laws that say to AT&T, as well as the national physical layer owning state body: "you can operate this cable, but you must treat all the bandwidth buyers equally."

Regrettably, even though privatizations are in progress in many areas, recent policy statements by the Ethiopian government have indicated neither a move toward deregulation in the telecommunications sector, nor one toward separation of the operator and regulator functions of the government in Telecommunications. To the contrary, recent actions have been in the opposite direction, with the only existing e-mail access provider in the country on the verge of having its service taken over by the Telecommunications Authority, and a rumored exclusive deal with a foreign company to provide Internet connectivity of some kind.

Let us hope that the national Telecom monopolies in Ethiopia will successfully transform themselves from sole players into detached, open-minded and farsighted referees.

Back to top

REFERENCES

1. Anderson, C., "The Accidental Superhighway: a survey of the Internet", The Economist, July 1st, 1995.
2. Bertsekas, D., and Gallager, R., Data Networks, Prentice Hall, 1993.
3. Davies, A., "The Politics of Telecommunications: the decentralized alternative", Pinter Publishers/St. Martin's Press, 1994.
4. Kupfer, A., "Craig McCaw sees an Internet in the Sky", Fortune, May 27, 1996.
5. Maraa, W.C., and Schesser, J., "Africa ONE: The Africa Optical Network", IEEE Communications Magazine, Feb. 1996.
6. Miller, M.S., Drexler, K.E., "Markets and Computation: Agoric Open Systems", in The Ecology of Computation, Bernardo Huberman editor, Elsevier Science Publishers/North-Holland, 1988.
7. Notebaert, R., "Technology's Contribution to the Competitive Marketplace", IEEE Communications Magazine, Dec. 1995.
8. NRenaissance Committee, "Realizing the Information Future: The Internet and Beyond", National Academy Press, 1994.
9. Semret, N., A Resource Allocation Game with Application to Wireless Spectrum, submitted for publication, 1996.
10. Tsigie, A., and Feyissa, G., "Telecommunications in Ethiopia: Past, Present and Future", in Telecommunications in Africa, E. Noam editor, forthcoming.

Back to top

About the Author: Nemo Semret is a PhD candidate at the Centre for Telecommunications Research, Columbia University in New York. He can be reached at nemo@ctr.columbia.edu

Back to Web Resources