Eritrea has one of the lowest per-capita income levels in the world, variably estimated between $150 and $250 per person per year. The literacy rate is about 20%, and the local languages are Arabic and Tigrigna which have scripts that are not easy to use on computers as yet. The scarce resources means that there are few computers (perhaps 10,000 to 30,000 in the country---less than one for every 100 people), few people can use them, and there little money to pay for Internet services. Secondly, by having a small population of about 3 to 4 million people, and a small computer market, there is not the demand necessary to have a large Internet services market that allows providers to develop economies of scale. And third, with Internet services being potentially expensive, and with the country short of resources, the government is hestitant to subsidize advanced communications infrastructure that will be used by very few Eritreans. The Eritrean government wants to make sure that Internet development contributes to the national development process rather than using up scarce technical and financial resources.
As a result of the above-mentioned challenges, a favorable policy and economic context for Internet services is just not available. The markets don't have much money or demand for Internet services, and the government is justifiably skeptical of the national benefits of Internet development.
But in spite of questions regarding the true benefits of Internet, at the beginning of 1999, the Eritrean Communications department solicited license applications from local businesses for Internet service provision. In addition, in August 1999, the Eritrean government signed a memorandum of understanding with the U.S. Agency for International Development regarding assistance for the establishment of Internet in Eritrea through the Leland Initiative. But even with official sanction and support of Internet development, there are many details and difficulties with regards to actually bringing effective services to Eritreans. This article discusses some of these more detailed issues and challenges.
In order to properly bring Internet connectivity to Eritrea it is necessary to first be clear on the reasons for Internet connectivity and Internet services.
For many Internet users there is no question that the Internet is a good thing. It is not even a debateable issue for them. The average information addicted netizen, will say "Of course the Internet is good for Eritrea...It brings information, and information is power and money...which Eritrea needs!" While for many in Eritrea the question seems irrelevant: 95% of Eritreans have no computer access, and even if they had computer access they may never afford Internet access. So who cares?
The assumption of the analysis in this paper, is that there are three requirements which if met, will make Internet services feasible and highly desireable in Eritrea.
The first requirement is that Internet is a good investment: That is if makes good business sense and makes money without subsidy, then it is just plain good business. If Internet services make money on their own without any subsidy, then they are contributing to Eritrea's economic growth and development.
The second requirement is if the Internet is accessible to almost any Eritrean with access to a computer and a modem no matter what their economic level, then it is good for Eritrea. If information resources are more available to Eritreans, then it decreases the monopoly power that foreigners and foreign institutions have on information and information resources. This makes Eritreans and Eritrean businesses more competitive relative to their foreign counterparts.
The third requirement for Internet in Eritrea is that it strikes an optimal balance between price and performance for the average user--delivering reasonably good service at a low price if possible.
We assume that if these three requirements are met, then there will be a consensus that Internet connectivity is good for Eritrea.
(There are other concerns like people accessing objectionable or pornographic content. These problems are very easy to solve technically by setting access restrictions at the international connection. Policy is more of the question with those issues)
We of the Eritrea Technical Exchange (a non-profit support project in California, USA), wish to present for Eritrea and Eritreans, some of the details of how the Internet can and will be brought to Eritrea. The ETE is working to assist Internet connectivity in a way that will completely fulfill the three requirements and resolve most questions and concerns about Internet access in Eritrea.
The main innovations of the activities that we are supporting include the the following:
Our organizational model for such email development combines foreign volunteer technical support, and local commercial operation of the system by national businesses. A second component of the model, has been to use--at least initially--some margin from the commercial sale of email services to subsidize access for public or selected governmental institutions.
At first, there were doubts of whether or not email was good or even legal in Eritrea. But the success and efficiency of the email system in practice has resolved these doubts. Some of this involved showing people throughout the public sector that email was a good and useful service. This was done by providing people with free email accounts for a period so that they could see its benefits.
Like with email, Internet-mediated information access and distribution is several times more efficient and cost effective for certain types of applications. And once Internet is demonstrated in a way that is sensitive to the national interests of Eritrea, there will be consensus that Internet is good for the country.
Furthermore, Eritrea may take pause from Ethiopia's experience with Internet development. The service there was soon oversubscribed, and it has been plagued by slow performance, high cost, and low efficiency.
The Ethiopian solution to the Internet access problem is far from optimal. The reason is that average price/performance is poor, and there is only a few price/performance options in an economically very diverse market. The solution is to increase the overall performance of the system (through efficiency improvements), and to provide a wide economic range of Internet use options for the local market. These solutions can and should be included in Eritrea's implementation of Internet.
The Internet is a type of Wide Area Network or WAN. What a WAN does is deliver packets of information or data to potentially distant computers over a wide range of interconnected computer networks. These networks may be of widely different types. But technical standards and rules have been developed for the delivery of such packets. These standards make the efficient operation of WAN's possible. The standard or protocol used on the Internet is called TCP/IP which stands for the Transmission Control Protocol/ Internet Protocol.
Computers can be either directly or indirectly connected to the Internet. Indirectly connected computers use an intermediary, or "proxy" computer to get data from the Internet on their behalf. Directly connected computers must have a unique IP (or Internet Protocol) address which is a series of four numbers, each between 0 and 255 noninclusive.
National Internet connections must have at a minimum two components. One is the local Internet Service Provider that provides the different servers and computers that the users use for calling in, storing and retrieving email, hosting web sites, etc. And the other is the Network Backbone Provider. This is the institution that provides the international data connection, and central network management.
For Eritrea, the international connection can come in several forms. One form may be an international leased line such as a 64 kilobit/second voice channel, or a VSAT connection. Alternatively the International connection can be provided through a simple dial-up connection using regular phone lines, or a roaming dialup connection through an international ISP that provides such roaming services (such as iPass).
|Data Transfer Method||
Cost per Minute
Cost per Megabyte
$1 - $2
$10 - $20
$0.25 - $0.5
$4 - $10
|Private VSAT (128 kbps)||
$0.20 - $0.40
$0.5 - $2
|Lease Voice Channel (64 kbps)||
$0.05 - $0.10
$0.4 - $1
In calculating these costs, we make the following assumptions. For phone lines, we have about a 2.5 kilobyte per second transfer rate at $1.50 per minute, which leads to 400 seconds or 6.7 minutes of transfer which is a little over $10 per Megabyte. Call set-up and other inefficiencies can increase this to $20 per Megabyte.
For international roaming services we assume a low quality connection of about 1 kilobyte per second (or less) which means 1000 seconds or 17 minutes of transfer is needed which is $4 at the low end and up to $10 per Megabyte if there are other inefficiencies such as a degraded connection.
For private VSAT, the cost would be $10,000 to $20,000 per month for a 128 kbps private connection typically. This is about $0.23 per minute. But at most such a connection would be used at 30% of capacity so we would get a transfer rate of 1000 Kilobytes /[(128 kbps/8bits/byte) X 30%] = 200 seconds or 3.3 minutes/Megabyte transfered. This means that for a well-utilized line the cost would be about $0.6 per Megabyte transfered.
For a leased international voice channel, it is assumed that such a line would be provided on a cost recovery basis by the telecom authority. With such an assumption the line would be about $2000/month, or about $0.05 per minute. A 64 kbps line can download at about 8 kilobytes per second but for a leased connection it would likely be used at less than 30% capacity so it would take 1000 Kilobytes/(8 kilobytes/sec X 30%) = 400 sec = 7 minutes or $0.35 per Megabyte.
So it can be seen that depending on the type of international data connections that is available, and the efficiency with which it is used, the cost of transfering data from the Internet to Eritrea can vary by a factor of 50.
Currently, for the email system, the bandwidth costs account for approximately 30% of the retail price of the email services in Eritrea. Other costs are effectively the technical service, capital recovery, and administrative operational costs of system administration and service provision.
Meanwhile web-browsing can make significant demands on the network. And a single page of information may be 0.1 Megabytes or more. Furthermore most people will not want to wait for more than half a minute to minute for the information to be displayed on their screens. Only 200 people downloading one page each can conceivably require as much data transfer as the entire country's email system in one day. And each user may wish to have their data transfered at a rate of 2-3 kilobytes per second. This can make tremendous demands on a network which has a limited international connection.
For international web browsing over a limited international connection, one must have either very few users, or very slow data transfer (i.e. it may take several minutes for a page with a lot of graphics to be displayed). This is the fundamental delemma of Internet access in remote locations with limited telcommunications infrastructure.
Eritrea's computer communications market is diverse. At one extreme are international business people who need to connect at the Internet at speeds which are only limited by the modem in their laptop. At the other end are poor students for whom 50 Nacfa (US$ 10) per month is a big expense.
For business people, their alternative is making an international phone call to Europe or the U.S. and connecting to a service provider there. Such a call would cost $2-$3 per minute originating from Eritrea and $1-$2 per minute originating from outside Eritrea. On such a connection, a user can obtain transfer speeds of 2-4 kilobytes/sec. Their main requirement of an Eritrean Internet service, would be U.S.-level service at a price that is cheaper than the alternative of making an international call to a foreign ISP. If our target price is 1/2 the cost of an international dial-up connection, then $0.50 to $1.00 per minute is a potentially acceptable price for top quality Internet access.
For the poor students, their main market demand is acceptable access at rock bottom prices. They would probably want at least a few hours of Internet use for their hard-earned 50 Nacfa. If we say this corresponds to a little over 10 minutes per day, then we have 300 minutes for $6, or $0.02 per minute of use. And if they get a few hundred pages of information per month, they will probably be pretty happy.
Therefore, the 'ability to pay' for the different components of the Eritrean market varies by a factor by more than an order of magnitude. This is simply a reflection of the fact that per-capita incomes in Eritrea are less than 1/100-th what they are in the wealthier, more technologically developed countries. Ideally, an Eritrean Internet can satisfy the different demands of both expatriot and local users.
There will be two main private sector impediments to cheap email-only accounts. The first and most important will be that supporting such services will compete with full-service Internet accounts for the technical support resources within local ISP's. The overhead and profit margins on the full-service accounts will be higher than for low-end e-mail only accounts. Given this fact, local ISP's may or may not offer such services even if it is strictly economical to do so.
Especially for local ISP's that make the commitment to a leased data line, their primary objective will be to develop a revenue stream that can support the continuing cost of a leased line. If they can charge $20 - $30 for a full-service connection, they will put a premium on making sure there are enough full service accounts to provide their revenue stream. It therefore may be in their interest to shut-out low cost email accounts even if 75% of the potential low-cost e-mail users can't afford a full service account. The key factor in whether or not very low cost e-mail access is provided in Eritrea will depend on there being enough competition between ISP's so that at least one of them finds it profitable to provide services to the low-cost email market.
Over the long term it will be in the interest of local ISP's to provide some very low cost access services and to provide free access services to students in educational institutions. Market development in Eritrea for computer communications services is growing exponentially with a doubling time of 6 months to a year. Providing low-cost entry into this market will allow more people to develop experience in computer network use. The lack of experience by local users in computer networks is the primary limitation for local demand development. This is because with its very low level of computerization, and its reliance on slower paper-based organization, most Eritrean organizations can potentially reap very economical benefits from increased levels of computerization. The lack of skilled personnel is the primary impediment.
Therefore, given rapid, exponential growth of computer comunications markets (even in Eritrea), small increases in market expansion now will have exponentially large impacts several years down the line.
Hyper-Text Mark-up Language (HTML) is rapidly becoming a platform-independent standard for formatted documents on the Internet, and this language has multilingual utilities built into its standards. HTML is the computer language in which most Internet content is written, and most e-mail programs are now capable of writing and displaying messages with HTML formatting. In HTML there is a font specification command that sets the font or script in which the message is displayed. This is the method by which email messages are sent in Chinese, Thai, and a great variety of other languages. The exact same methods apply to sending messages and producing web content in Tigrigna and Arabic.
In addition to the fonts, one needs a method for typing the script into the computer. This involves the use of a program that remaps the keyboard to produce the proper letter encodings corresponding to what is being typed in by the user. Several programs for doing this in Windows are available in Eritrea and Ethiopia, including Yada (by Tfanus Enterprises) Ge'ez Gate and a host of others. Software for composing messages costs a significant amount ($90 for Yada for example), but the fonts for displaying the result are free. This means that it is currently possible and perfectly feasible to produce local-language web content. Web sites that have this content, can also provide an FTP (file transfer protocol) link for downloading the fonts so that anyone can view the content for free.
Local language email and web content has a couple of national advantages for Eritrea. The first advantage is that it makes the information in local languages more accessible to Eritreans who speak and read primarily Arabic or Tigrigna. This can dramatically expand the relevance of Internet services.
The second national advantage of local language Internet content, is that it decreases the competative advantage of foreigners with regards to utilizing locally produced information content. One disadvantage of producing content for the Internet is that the result is often that the information is more accessible to someone abroad than it is for someone in the country in which it was produced. On the other hand, reports in Tigrigna and Arabic will be easier to for Eritreans to write than English reports. And also, once posted on the web, it will be far easier for Eritreans to read and interpret these documents. Foreigners who want to utilize this information will likely have to hire Eritreans to translate and interpret the documents for them.
A near-zero marginal cost of web content for these accounts is
attained by limitting content to what has been retrieved by other users
at other times. The storage cost of this content is nearly zero since
the cost of a 10 Gigabyte hard drive is only about $200 or $0.20 per Megabyte.
This per-megabyte storage cost is very small compared to the cost of transferring
the data over the international data lines.
Several government organs are responsible for providing information services to other ministries, local governments, and the public. Some of these include: Meteorology services who are responsible for weather reports and forecasts to other ministries and agencies, the Department of Energy which is responsible for energy and electricity system standards, the Ministry of Agriculture which is responsible for agricultural forecasting and the distribution of agricultural inputs and extension support services, and the Ministry of Construction which is responsible for managing and approving construction projects. All of these agencies could use local Intranet services to provide an alternative means of publishing reports and updates. In many cases, publishing documents on the Eritrean Intranet or making them available for download will be cheaper and easier for many types of constituents. Many documents are prepared on computer (often in MS Word), and a site that allows such documents to be downloaded can be constructed in a matter of minutes on an existing server.
Also, if the Internet becomes popular amongst Eritrean ministries and both public and private organizations, certain information such as weather reports, forecasts, and crop status which is useful for a wide range of productive institutions, could be more easily obtained by surfing the local web than by visiting the relevant ministry and asking for a printed report which may be expensive. National Intranet distribution of some information may be more effective than current paper distribution methods in some cases.
Another application of local Intranet services is to allow local organizations to engage in networked data collection and information administration activities. A dial-up line to an organizational data entry and retrieval interface could be useful for a wide range of local organizations. This is another service that can be more fully developed with local Intranet development.
A low speed connection with download speeds of about 0.5 kilobytes per second will cost at least $0.25 per minute while a higher speed connection with download speeds of over 4 kilobytes per second will cost over $1 per minute. In addition, 24 hour-per-day real-time browsing will be one of the last services to arrive since it is one of the most expensive services to support. But for the average user 50% to 80% of web content will already be available on a local proxy-server, so for many users there will the the 'appearance' of real-time Internet access even when the international Internet connection is congested.
Internet and national Intranet services are perfectly capable of providing multi-lingual, voice, sound, and video publication and distribution services at reasonable cost. Which of many possible services are actually marketable in the Eritrean context remains to be seen.
Just to cite on example, it would be perfectly feasible to send voice
recordings internationally as email attachments. For example when efficiently
compressed, voice content requires only 8 kilobytes per second of sound.
This means that one minute of recorded voice can be sent in 480 kilobytes.
This means than when the per Megabyte cost of data transfers drops below
about $4, it becomes more economical to send voice recordings via email
or Internet than it does to send then over phone lines. This opens
up the potential of sending voice recordings over the Eritrean Internet.
And if done right, this is a service that all Eritreans could enjoy.
It should be noted that other technical solutions besides what is being described here exist. In particular, both Microsoft and Sun Microsystems provide commercial computer networking solutions for ISP's. But given the Eritrean desire for self sufficiency and the need to preserve precious foreign exchange dollars, it is likely that the free software solutions that the ETE is promoting and developing for use with inexpensive PC's will out-compete standard commercial solutions based on Windows NT or Sun servers.
Linux is free, and installs readily on any personal computer with a 386 processor or above. Recently, a fairly user friendly release of Linux called Redhat has become popular and allows one to install the system fairly simply. For more detailed information on the Linux operating system the reader should refer to more definitive websites such as http://www.linux.org/
A hunt group is a collection of several phone lines that can be accessed with one phone number. At the telephone company, the switch is configured to attempt to connect to the phone lines in sequence. If the first number is busy the switch automatically switches the call to the next available line. Hunt group phone line configuations are currently available in Eritrea and are currently being used by local ISP's for dial-up access. The hunt group configuration allows fewer phone lines to be used at greater average capacity for call-in users, using existing phone infrastructure at greater capacity and efficiency.
The current volume of computer communications customers in Eritrea is still in the hundreds. The number may double every six months to a year, meaning that for the next several years the number of customers will remain below a few thousand. So assuming that up to 1,000 customers may use one ISP and about half of these use the connection for web surfing in addition to email, we can calculate the approximate number of phone lines required. If each web-surfing customer uses the service an average of half an hour during an eight hour day (this low use volume per customer can be enforced by having a significant connection time fee) then an ISP might need as much as 30 phone lines. Over the long-term a substantial capacity for running multiple phone lines into Eritrean ISP's will need to be developed.
At earlier stages of development, ISP's can configure a server with 4 to 32 dial-in phone lines by using a multiport serial card that can connect to a bank of external modems (see for example: http://www.comtrol.com/sales/specs/rocket.htm ). The modem that we have had the best experience with in Eritrea is the U.S. Robotics Courier V.Everything modem. The larger, more expensive (about $200) V.Everything modems appear to have better performance in the face of long international line latency, high phone noise levels, and phone line unreliability. When the number of phone lines coming into an ISP approaches several dozen, then more advanced technologies for bringing bundles of phone lines (or digital phone lines) may need to be investigated. But for the time being., multiport serial cards and individual external modems are easy to implement and can be customized in their operational qualities and parameters in Linux.
Some of the features of a PPP 'daemon' program that allow for more efficient
utilization of dial-up connections includes on-demand features that allow
the server to make an international network connection only when it is
actually needed by an ISP's customers or a business/government institution,
a configurable activity time-out, and the ability to run completely customized
scripts at the beginning and end of the connection session. Customized
connection establishment and termination scripts allow for complete control
and setting up logging and billing systems along with access restrictions
for different types of user services.
There are probably half a dozen other possible approaches that can be taken local language content development. A diversity of approaches should be tried with the key to success probably be being some standardization of the fonts, key mappings, and transliterations, along with some low-cost access or entry into multi-lingual computer use that does not require the purchase of some proprietary software. Those two elements will allow for wide-spread multi-lingual access.
Development of local content requires allowing a diversity of users and developers to post content on their own web sites and providing an easy-to-use interface to allow people to do such web content development. In terms of increasing the customer base, and the local demand of Internet services. It actually might be to the benefit of local ISP's to provide automated web hosting services for their users for little or no extra charge for Internet user accounts. Even more beneficial for local ISP's would be to organize contests amongst their users so that those with the most visits receive some reward like free Internet services for one to several months. By providing a user friendly interface for local users to upload web content, one allows the creation of a pool of potentially hundreds of Eritrean web authors rather than just utilizing an elite of a few technical staff working at the local ISP's.
Internet services will arrive in Eritrea in the fall of 1999, shortly before the new millenium. It is uncertain if the Eritrean branch of the International computer internetwork will be the elite domain of computer saavy foreigners or an accessible tool for national economic development. For friends of Eritrea, it is certainly important for Internet and computer information services to be accessible to as many Eritreans as possible at very cheap prices. But the cost of data services depends not only on an inexpensive international connection, but it also depends on the ingenuity in innovations of Eritrean ISP's in managing data transfers, local hosting of foreign content, and the development and expansion of local content. How much ISP's can direct users to content that is already present on the hard disk of a local server will determine the cost of Internet access for Eritreans as much if not more than the cost of international connectivity.
The degree to which the Internet becomes relevant and accessible for
Eritreans will depend on the ingenuity, inspiration and dedication of both
Eritrea's friends and computer and business professionals in Eritrea.
The ETE hopes to contribute to information accessibility for Eritreans.
Information access must not become the monopoly domain of a few affluent
foreigners. The Internet is a public resource, and the public in
every country and at every income level should have the right and the means
to access it.
Most recent update: August, 1999 by rvb