While Vice President Al Gore promotes the inclusion of Africa on the Internet through the multimillion dollar Leland initiative, the question remains as to what extent such computer communications developments are desireable or feasible in societies with more pressing problems of basic human health and survival. The vision of a seamless world with people of all cultures and nations communicating freely has much appeal. But in Africa this attractive ideal meets some hard realities where the Internet might actually benefit foreigners in Africa rather than Africans themselves.
In this article, we present our experience with providing reliable email access in one of Africa's 52 countries and discuss some of the technical and social issues we encounter.
Eritrea is Africa's newest country. It won its independence after a brutal 30 year liberation war with Ethiopia which ended in 1991 with the victory of the Eritrean People's Liberation Front (EPLF). A national referendum on independence was held in 1993 with 99.8% of voters favoring independence from a technically more advanced Ethiopia. Eritreans then set about the task of national reconstruction in a country with a per capita income below $150 per annum, an illiteracy rate of 80%, a life expectancy of about 50 years, and with 80% of the population without even basic electricity.
Since computer-based communications are by far cheaper and more efficient than letters and fax, it is necessary to try to provide access to this form of information exchange to the poorest countries and to examine in which ways they might benefit. To do otherwise is to force the poor to pay much more to communicate internationally than technically advanced countries.
Upon independence there was no reliable email or computer communications services even in the relatively advanced capital Asmara, though unreliable links were occasionally established with services in Ethiopia. We began work on Eritrean email service development while one of us was teaching Physics in 1995 at the University of Asmara, the national university. At that time there had been several abortive attempts at establishing basic email services. These attempts had failed largely due to organizational and infrastructural problems that resulted in unreliable connections or broken equipment.
Organizational and material infrastructure in Eritrea is weak at best. Power outages occur more than once per week, there is cross talk and a high level of noise on the phone lines (local wires are parallel strand rather than twisted pair and the local radio station can often be heard), what few computer technicians exist are swamped by the large demand for basic repair and training services, and equipment lifetimes are shortened by dust, lightning and fluctuating voltage supplies. The challenges facing computer communications development in such a setting can be summed up in three words: Accessibility, Reliability, and Sustainability.
In February 1995, a small non-profit project: the Eritrea Technical Exchange launched an effort at email services establishment in Eritrea that has evolved over the last two years into a national dial-up email system. Because of very poor phone line quality, approximately six months of part-time experimentation was necessary before a reliable but expensive email transfer system (approximately $0.50 per message) existed. After establishment of a reliable DOS-based system using UUCP for email transfer, implementation of an in-country client-server system using a Linux server for batching and compression of email improved through-put by a factor of 5-10. This reduced costs enough to expand email services to paying customers at the beginning of 1996. Such paying customers (private businesses and foreign individuals) provided sufficient income to subsidize project operation and expansion for educational and government sectors. The initial successes of the system sparked interest in the private sector which then established independent servers using the same methods. Later it was found that amplification of the modem output signals on the phone line greatly increased the reliability of the connections and increased through-put by another factor of 3-5. The resulting email system has operational costs for international email transfers below $10 per megabyte of uncompressed text. In spite of nearly no in-country UNIX system administration experience, email servers have been established at the national university, two private businesses and one government office. The system is serving nearly 100 institutional and individual connections. One year after the initiation of the operational phase, email volumes are approximately 300 messages/day exchanged or 30 megabytes/month uncompressed.
As with any `development' project, questions of social and political consequences need to be raised. Just because the Internet is the latest rage in the U.S. and Europe does not mean apriori that it is good for Africa. Who benefits from developments in computer communications? Is it a broad middle class or an elite few?
The stark reality of computer communications in Africa is that it is and will remain accessible to only a select few. Currently, only a tiny fraction of Eritreans can possibly benefit from computer communications or email, by virtue of the great limits on computer access. Simplified estimates of the total number of computers in the country yield an amount less than 10,000, or one computer for every 300 people. Assuming that each computer serves three people or less on average, this indicates that less than 1% of Eritrea's three million people have computer access. So any developments in email access will necessarily benefit only the most elite or privileged sectors. But in a country as poor as Eritrea, this elite includes students, university staff, and government workers who typically earn salaries ranging from $50 to $200 per month (4-16 times the per capita average). Such email users hence represent the poorest of those with email access internationally.
Those who most fervently demand email services in Eritrea are foreign expatriots who have had experience with the Internet abroad. Next comes foreign charities or non-governmental organizations (NGO's), and businesses. Finally, a fairly sizeable number of government offices and departments of the national university also request email access to aid in their international communications. Foreigners tend to have had some training or experience in internet use and email and tend to utilize the medium more voluminously than their Eritrean counterparts. This is reflected in the use rates where we have observed nearly six times more volume from foreign individuals (average use: 85 messages/month/node) compared to Eritrean government workers (use: 15 messages/month/node). This is partially due to the the conditions of scarcity which Eritreans have experienced, combined with unfamiliarity with the English language and typing. Exceptions to low Eritrean email use rates can be found at the University where academic staff have developed their ability and enthusiasm with electronic mail, where people's English-language and typing skills are good to excellent, and where each computer is used by a larger number of people (sometimes over 10 mailboxes on one machine).
In the University especially, email has increased the ability of Eritreans to organize international cooperation and exchange. The low cost of such communications and rapid turn-around times have enabled the organization of scholarship opportunities that otherwise would not have been possible. Students and workers in education have shown the greatest enthusiasm for computer communications and may provide the best area for more popular access.
Email, being a well-understood and proven technology, should have been simple and straightforward to configure and implement. But in spite of our expectations, we found in Eritrea this was not the case. International telephone line quality to the locations of interest was so degraded that even the simplest modem modulations (V.22bis or 2400 baud) could not initially be maintained, and a careful optimization of international file transfer methods had to be performed. Furthermore, because of the shortage of phone lines for server use, simple and robust methods for performing file transfers across local lines utilizing a minimum of line time had to also be developed.
Given the weaknesses of the local infrastructure we decided to begin our efforts with simple PC-based email programs and methods utilizing shareware when possible, yet maintaining UNIX-compatibility. We therefore selected UUPC (a shareware version of uucp for PC's) which we loaded on a 286 laptop and linked with a server in San Francisco, California over regular phone lines and a Telebit T1000 modem. The initial experiments were done from the Computer Center at the University of Asmara over noisy dial-up lines. In spite of some initial successes it was found that the noise levels in the phone lines (possibly combined with the intercontinental latency) made it impossible to maintain reliable connections even with 2400 baud connections. In order to maintain reliable connections it was found necessary to use Telebit's Packetized Ensemble Protocol (PEP) combined with specialized register settings that decreased the number of bits per channel used in the transmission. Telebit customer support was instrumental in optimizing this configuration. But even with PEP, throughput was approximately 200 bytes/sec during file transfers while large time delays of 5-15 seconds were required by the UUCP protocol to set up file transfer. For transfer of mail with UUCP, two files are sent, a control file (with the rmail command) and a data file (with the message contents). The per-file time delay combined with the slow tranfer rates resulted in mail transfer rates of only 1-3 messages/minute for this initial phase of system development.
These slow transfer rates were partially ameliorated by switching to batched compressed transfers. To do this, the server in Eritrea was upgraded to a Linux operating system runing Smail as the mail transfer agent. With Smail it was simple to incorporate a compressed batched transport which consisted of placing mail files in an outgoing queue directory, concatenating and compressing the files according to a prearranged schedule, and queuing an outgoing UUCP transfer with uux. At the receiving machine the uux job uncompresses the batched mail and feeds it to Smail for delivery.
The transfer rates for the batched compressed mail system after January 1996, have gone through several stages of improvement. During the initial phase, Telebit PEP protocols were being used with an unamplified output signal on noisy phone lines. The first improvement is due to an amplification of the output signal on the Eritrean side of the connection and a switch to V.32bis (14.4 kbps) modulation. But these tranfers still lacked complete reliability which was affected by the phone line `weather'. Noise and transmission levels in the phone lines would be impacted by affects such as the volume of city-wide phone traffic and the broadcasting time of the local radio station (causing night-time transfers to be more reliable than daytime transfers). File transfer rate and reliability was further improved by also amplifying the output signal on the U.S. side of the connection. This increased mean transfer rates to a steady 700 bytes/sec [note: calculate real number] for compressed files which were limited to 40 kilobytes maximum size. Further improvements were made by increasing the size of individual files transfered and by economies of large file-transfers brought about by increased email volume. Further improvements in transfer rates are planned with use of amplified V.34 modulations or the installation of phone line repeaters.
The current transfer rates are running at approximately 1000 bytes/sec for large compressed files. Dial-up line costs are about $1/minute (+-15%) implying a transfer cost limit of $18 per compressed megabyte. We feel that with amplified 28.8 bps connections further improvements to about $10 per compressed megabyte (or $3 per uncompressed megabyte with a typical compression ration of 3:1) can be obtained, but that this is the lower limit for transfers over dialup lines from the U.S. to Africa.
It should be noted that such international information transfer costs are far cheaper than any other form of international communication available to people in Eritrea. International fax transmission costs $1 - $2 per page while an international airmail letter costs $0.50 for one to five pages. By comparison, one page of text (approximatly three kilobytes) will cost $0.02 in pure transfer costs (overhead and maintenance probably doubles this cost to $0.05 per email page) though the system user also must pay for incoming mail which doubles the effective cost to the user to $0.10 per page sent, assuming equitable email transfer patterns.
Dedicated Internet connections have much lower cost per kilobyte transfered than dial-up connections. A dedicated 64 kilobaud satellite link to Eritrea would cost somewhere in the range of $3000 to $5000 per month. Such a connection has a transfer cost of $0.0002 per kilobyte if fully utilized, and would equal current transfer costs if monthly data volume reached 400 megabytes per month. Reaching these data transfer volumes within a few years will require an expansion of the computer communications market perhaps through the introduction of public-access computers, development of communications services in the local languages (Arabic and Tigrigna), or developing dial-up non-email Internet services ... perhaps a local www service with a local distributed database which later develops connections to the Internet.
Developing non-email Internet services for foreigners and NGO's could generate potentially high revenues. Currently such organizations must pay $2-$3 per minute for connect time to ISP's outside the country and several organizations are paying such prices for Internet connectivity. The availability of Internet services would provide a market for high-fee expanded-service connections. But such connections would require higher performance computers and a higher level of financial resources from users. On the other hand expansion of data communications markets by providing greater computer access and collecting smaller income from a larger, poorer market would serve the social goal of providing more Eritreans access to low international data transfer costs (currently less than 0.02% of Eritreans have access). Basic communication services including in-country and Tigrigna language access may be of interest to a sizeable sector of the population (which do not use the roman alphabet but Ge'ez or Ethiopic script). The commercial email providers in Eritrea are making attempts at providing at least limited email services in local languages.
Presently independent satellite communications outside the national telecommunications authority are not permitted. But there exist plans to establish an Internet link through the central government Eritrean Information Services Agency (EISA). This effort is being subsidized by the Leland Initiative. The details of how this connection is organized, utilized and managed remains to be defined. But the existing email system provides an incipient computer communications infrastructure with which the government-sponsored Internet connection will interact.
Whether or not particular computer communications systems are technically and economically sustainable is an important question in the African context.
In this regard, there is a direct trade-off between technical sustainability and capability. Usually the more capable and powerful systems are also those which are more complex and require a more technologically advanced infrastructure to maintain them. Such complex technologies will have greater capabilities but may be less sustainable because of the lack of local resources to support them.
Meanwhile on the ecomonic front, markets in Africa are cash-poor and it may not be easy to recover expenses for computer communication services if costs do not match marketability.
The main technical limitation to expansion of computer communication services in Eritrea is the establishment and maintenance of a network of servers that can provide users and institutions with dial-up access. Up until 1996, there were no reliable email or Internet service providers in Eritrea, and this remains the main bottleneck for the expansion of computer communications. With purely DOS or Windows based systems, server establishment and maintenance is not a problem, but such systems lack the reliability, flexibility and robust multi-tasking capabilities of real operating systems. We have found the local maintenance, propogation and operation of PC-based Linux servers difficult but possible in the Eritrean context. In spite of the existence of fewer than a handful of computer personnel in the country with UNIX system administration experience, several local business have had sufficient interest in providing electronic mail services to make the appropriate investments in hardware and training. For overloaded computer services businesses in Eritrea, technical personnel can dedicate only a relatively small fraction of their time to email systems development and operation. We provided free installation and techical support for businesses wishing to establish electronic mail services. With the availability of such support, developing the technical capacity to provide reliable email services at an Eritrean business requires approximately two months of part-time work. During these two months the technical personnel develop a familiarity with Linux, email software, system configuration and monitoring, and client node installation procedures. In the first year of the national email system operation, two local businesses have established servers and are providing email services. Other businesses have also expressed an interest in server establishment and probably will also develop services in the next year. Furthermore, selected University students have learned aspects of the operation and maintenance of email systems in Eritrea. As a result, approximately seven technical personnel are actively involved in email services in Eritrea presently, while a year ago, there was only one.
In spite of the fairly smooth operation of the email servers there are persistent but surmountable operation and maintenance problems faced by the system. The most prevalent problems include phone line malfunctions, lightning surges, and power outages. Serious phone line malfunctions have occured on all servers. Establishment of the University server was delayed 4-6 months because of delays in fixing the phone lines to the Computer Center. For the main governmental server, construction severed the telephone trunk line into the service area, and it took one week for service to be restored. The commercial server also suffered unavailable phone lines for several days. A back-up plan for email routing over alternate lines is a necessary component for maintaining reliability of service. In Asmara, phone lines are not sufficiently protected from lightning-generated voltage surges. During the short 1996 rainy season, over 13% of modems in the system were damaged by lightning-induced surges. Hence surge protection of the phone line at the user-end is an absolutely necessity. Furthermore, power outages are at least a weekly occurance. During the rainy season the frequency of power outages were nearly daily. The duration of such outages are in the range of 20 minutes to a few hours in general, though outages of a day or more did occur during reconfiguration of the distribution grid or after lightning damage to local transformers. The appropriate strategy for dealing with outages is the installation of high-capacity uninterruptible power supplies (UPS) at all servers.
Another persistent concern is the limited availability of phone and related telecommunications services. Telephone personnel unfamiliar with computer communications may delay installation of extension lines for weeks because of a unfamiliarity with the intended application. Businesses interested in providing email services may have to wait several months before being able to install additional phone lines. Three of the four servers have only one line for servicing users, and such servers are limited to a user base of about 40-50 before users will start encountering access difficulties from a busy phone line.
Meanwhile, after an initial subsidy, economic sustainability is nearly attained because of the high international data transfer rates. Commercial customers pay $20-$25/month for an account that allows them a maximum monthly transfer volume of 1 Megabyte uncompressed. The bare transfer cost of this amount of data is about $6, while overhead and contingencies approximately doubles the bare cost to $12. The remaining $8/Megabyte allows for a margin that can be used to subsidize governmental and educational users, and accumulate resources for equipment replacements. There remains an ongoing subsidy from the U.S. non-profit in the form of donated labor and equipment costs, but these subsidies are gradually being phased out.
Overall in spite of the difficulties, the current system is likely sustainable. There is now a critical mass of technical personnel trained and experienced in system administration and maintenance. Furthermore, infrastructural problems, though disruptive, can be countered through proper planning and implementation of redundant preventative measures. And finally, after initial subsidy and intervention from an outside non-profit organization, the system is becoming economically sustainable through the efforts of local commercial businesses. The economic viability and sustainability of the system is gradually improving as subsidies are gradually removed, and operational costs continue to decrease.
Computer communications are a cheaper and more efficient means of communication or data transfer than voice, letter, or fax. If we are to avoid the contradiction of having the poorest countries paying the highest international communications costs, access to such interchange needs to be expanded. In Africa's newest country, Eritrea, we have taken the first steps in establishing email services which are much in demand. While email is not new to Africa its accessibility, reliability, and sustainability are continuing concerns. In Eritrea, the limited amount of computers and phone lines in the country constrain accessibility of email to less than 1% of the population in the near-term. But within the fundamental material constraint of computer availability, it is possible to organize email services so that poverty, weak infrastructure, and lack of personnel are surmountable challenges. To do this we find it necessary to first optimize international dial-up data transfer through batched mail combined with software compression, and amplification of modem output. This allows data transfer costs to fall below $10 per megabyte of text. Then by separating users into paying and subsidized markets, it is possible to charge internationally comparable fees to paying customers ($20-$25/month) that are sufficient to subsidize access for those sectors unable to pay (many government and educational institutions). This provides greater access to the medium for Africans, especially in the university sector. But even with such equalizing measures in place, we find that the majority of email use is not by Africans, but European or American foreigners. Introduction of Internet services would likely benefit even more affluent sectors which can afford higher performance computers, and longer connect times to commercial ISP's.
To compensate for infrastructure weaknesses, it is necessary to install phone line surge suppression, to utilize uniterruptible power supplies and power conditioning for the servers, and to develop redundant dial-up links. Meanwhile, sustainability requires initial assistance to both private and public technical personnel in system set-up, configuration, and administration. To develop a relatively self-supporting minimum network of email servers and support personnel has required a two-year, part-time, in-country commitment by our small non-profit support project.
The Internet has the capacity to transport information much more efficiently than ever before. As such, developing African Internet services is now the focus of several computer-related development projects such as the vaunted Leland Initiative. But without slowly and arduously developing in-country computer communications infrastructure only a tiny group of foreigners and businesses will benefit from Internet expansion. In order to break out of the confines of elite email and Internet utilization, it is necessary to develop mechanisms of nationally-adapted access-oriented development to counter the tendency of computer systems to preferentially benefit affluent foreigners. We find that Eritrean access barriers are material, cultural, and educational. Without overcoming these impediments to popular computer communications access, African Internet connections may very well become off-ramps to nowhere on the information superhighway rather than the international networks of intercultural communication that are being sought.