Larry Press <firstname.lastname@example.org>, USA
Luis Rodríguez <email@example.com>, Venezuela
Governments and international organizations regularly gather economic and census data for use by administrators, policymakers, legislators, investors, and others. This paper discusses the extension of such work to the Internet, with emphasis on developing nations. The paper begins with a description of a survey of 23 academic networks with international connectivity (IP or Unix-to-Unix Copy [UUCP]) in 21 Latin American and Caribbean nations. After summarizing the results of that pilot survey, we discuss questionnaire revisions and survey problems, and conclude by proposing further steps toward a network census for developing nations.
Governments and international organizations regularly gather economic and census data for use by administrators, policymakers, legislators, investors, and others. Several of these surveys are relevant to telecommunication. For example, the International Telecommunication Union (ITU) reports numbers of telephone lines, telecommunication staff sizes, investment levels and plans, and many other statistics in an annual report . The World Bank  and United Nations Development Programme (UNDP)  also gather data on telecommunication infrastructure as part of more comprehensive economic and social surveys. UNDP is primarily concerned with developing nations, which are also the focus of our work.
Several organizations also gather Internet census data. For example, Lottor  periodically conducts an automated host count by "walking" the Domain Name System to count hosts and attempting to ping a randomly selected sample of those he finds. Quarterman  uses a random sample of the hosts Lottor discovers and distributes a questionnaire electronically. Landweber  compiles an international connectivity table regularly, but it is limited to noting what type of connectivity exists--IP, Unix-to-Unix Copy (UUCP), Bitnet, etc.
Pitkow  and Yahoo  have conducted commercially oriented surveys of self-selected World Wide Web users, and O'Reilly Associates  used traditional survey research techniques to estimate user counts and characteristics. These surveys focus primarily on IP-connected hosts in the United States, and they are geared toward consumer demographics.
As a first step toward a more detailed census of connectivity in developing nations, we drafted a questionnaire (Appendix A) and distributed it to administrators of 23 academic networks with international connectivity (IP or UUCP) in 21 Latin American and Caribbean nations. After a summary of the results of that pilot survey, we discuss questionnaire revisions and survey problems, and conclude by proposing further steps toward a network census for developing nations.
The questionnaire covered five general areas: coordinates, the network, users, user support, and success stories. The coordinates section identifies the network and the person responsible for the survey. The network section is intended to shed light on budget and charging policy and the technical characteristics of the network. The user section is geared toward learning about users and the hardware and services they have access to. Since we wish to emphasize users, we also included a section on their support. (While there are many constraints on the spread of networks in developing nations, we feel the most difficult to overcome is the lack of a large, widely distributed, demanding, well-trained user community ). The final section asks about successful applications, since these can be used as examples for others and in justifying network investment.
Table 1 shows basic information on the networks we surveyed and illustrates the variance in these networks. They range from a small UUCP network with 50 active users and dialup, UUCP connectivity at 2,400 bits/second to modern IP networks. Note also the discrepancy between the number of "active" users and "registered" users. This reflects the cost and difficulty of using networks in developing nations.
Table 2 shows the budgets of the networks and the way they are spent. The large percent devoted to international connectivity suggests that intranational and intraregional cooperation might result in overall savings.
Table 3 summarizes the services provided the network users. We would expect that over time, the proportion with e-mail-only service will fall, because of evolution from UUCP to IP connectivity. Note that users with access to an intranational shell or Point-to-Point Protocol (PPP) account may still be restricted to e-mail on international traffic as, for example, in the case of CENIAI in Cuba.
Table 4 shows the affiliations of the users of the networks surveyed. Because we only contacted academic/research networks, the academic figures are high. While the first network initiatives in developing nations were typically (though not always) academic, there are now growing numbers of commercial service providers. We expect predominantly academic networks to continue serving their sector; but to varying degrees, many are now offering or considering commercial service in order to sustain themselves. There will be different legal frameworks and educational subsidies in each nation, but the academic-commercial boundary will soften.
More detail and discussion of these results are presented in "Metodología para la Evaluación de Redes" . The full replies to these questionnaires as well as others gathered at the INET'95 Developing Nation Workshop may be found at http://som1.csudh.edu/fac/lpress/xx .
An extended version of the questionnaire was used in a study of Cuban networks [3, 4]. It delved into more detail on the four networks with international connectivity and their subnetworks with questions on mission, staff, general topology, central location vs. outlying centers, special purpose server applications, inter- and intranational data link and transport protocols, databases, custom software, other networks served, plans for the future, and traffic volumes.
Most of the questions added for Cuba focused on the network, but we would also add user and implication focus. Press  suggests five areas in which networks might assist developing nations--economic productivity, health, education, democracy, and quality of life--and we could add questions that focused on each of those areas. For example:
It is clear that answering such questions would require cooperation beyond the networking community. Social scientists or specialists in education, health care, and so forth would have to be involved in an impact-oriented survey.
Note that in addition to gathering answers to these questions, a network-impact survey should report them as percentages, for example, the percent of high-school teachers on the Net. It would also be interesting to assess the government role in encouraging networking. What percent of the budget of the network is underwritten by the government, foundations, international organizations, users, etc.?
While a longer questionnaire would paint a detailed picture, there is an obvious tradeoff between survey complexity and the time required to complete it. Furthermore, while we are focusing on networks with international connectivity, they provide gateway service for other subnetworks, and our contact people may not have good information on those connected networks. With the Cuban survey, repeated follow-up was needed to obtain subnetwork information.
Networks are heterogeneous, so they need different questions. For example, there is no need to ask about application servers when investigating UUCP networks. Even where the same question is meaningful for heterogeneous networks, the answers may not be comparable. For example, overhead and support budgets are spent on different services in a nation with an advanced IP network than in a nation just starting out with UUCP.
This problem is compounded by the inconsistent interpretation of questions. For example, in Table 1, it is clear that some networks dated their first operation from the time the project was begun and others from the time they established their current connectivity. In conventional surveys, this problem is addressed by including detailed instructions and explanations of questions or by using trained interviewers. A network-based survey can use a listserver for questionnaire recipients to overcome this problem.
There is also the possibility of inaccurate reporting--fudging up or down in order to influence policy, funding, or prestige, or due to lack of time spent gathering information. For some questions, this could be mitigated by independent validation, for example, by surveying users.
Unwillingness to cooperate is also a problem. Our reply rate was 61 percent, which can be seen as a half-full or half-empty cup. In addition to the time required to complete the questionnaire, there may be some fear of informing "competitors" about your network. (For example, some networks skipped the budget section). On the other hand, we are encouraged by the fact that most participants provided complete data, and companies do share competitively sensitive data in established surveys by trade associations and others. For example, software companies provide sales and cost figures for the Software Publishers Association market research. National representatives and ministries also participate in surveys by ITU, World Bank, and UNDP. The latter deal with complex issues such as those we have suggested in our user-oriented survey revisions. Participation is also facilitated by our ability to use the Net for coordination and follow-up. At some time in the future, user demographic information may have commercial value that could be used to partially offset survey cost (note 1).
As indicated above, the questionnaire should be revised. Network administrators should participate in this process, as they know what is possible to report and what information they would like to have, and they must ultimately spend the time to complete the questionnaire. A first step is, therefore, the expansion of our list of contacts in and beyond Latin America and the Caribbean and identifying network administrators willing to guide this work. They can coordinate using an e-mail list explicitly limited to discussing the questionnaire for revision and consistency.
With a revised questionnaire in hand, we would like to update the survey semiannually. The detailed results would be posted on the Internet, and summaries reported in print and electronically. We would also like to prepare a user-oriented questionnaire, both as an end in itself and to validate portions of the network survey. This would have to be administered with the cooperation of network administrators.
There are difficulties in conducting an Internet census, but the results would be useful in setting policy and satisfying our curiosity. The fact that there can be no perfect survey should not dissuade us from doing the best job we can.
International Link Operate Reg. Active Speed Leased/ Network Nation Since Users Users (Kbps) Switched RECyT Argentina 1990 1,117 624 384 L BOLNET Bolivia 1991 1,200 800 64 L REUNA Chile 1992 500 500 512 L RdC Chile 1991 2,500 2,500 256 L Red Cetcol Colombia 1994 10,000 128 L TELECOM-CO.Colombia 1994 700 400 64 L CENIAI Cuba 1982 732 732 14 S ECUANEX Ecuador 1991 202 165 19 S RAIN Nicaragua 1994 1,100 900 14 S CNCnet Paraguay 1993 60 50 2 S LEDNET Paraguay 1989 70 60 2 S RCP Peru 1991 7,716 6,900 128 L REDID Dom. Rep. 1992 120 100 10 S RAU Uruguay 1991 325 266 64 L REACCIUN Venezuela 1990 2,500 1,800 128 L
Budget Equip. & Comm. Overhead Network ($1,000) Staff Software Link Other RECyT 1,300 40% 15% 30% 15% BOLNET 48 40% 60% REUNA RdC 80 24% 20% 45% 10% Red Cetcol 2,000 Telecom-Co.net CENIAI 41% 17% 28% 14% ECUANEX 100 50% 20% 10% 20% RAIN CNCnet 45 55% 20% 25% LEDNET 37 73% 16% 11% RCP 98 57% 55% 17% 4% REDID 10 100% RAU 250 18% 28% 53% REACCIUN 1,300 15% 20% 50% 15%
Email Dial Dial Direct Network Only Shell SLIP/PPP IP RECyT 82% 2% 1% 15% BOLNET 100% REUNA 8% 80% 12% RdC 85% 15% Red Cetcol TELECOM-CO.NET 2% 98% CENIAI 40% 60% ECUANEX 90% 10% RAIN 2% 15% 3% 80% CNCnet 100% LEDNET 100% RCP 65% 10% 10% 5% REDID 100% RAU 68% 6% 26% REACCIUN 100% 15% 1%
Univ. & Research Intl. Staff Students Govt. Commerce NGO Org. Other Network RECyT 48% 10% 18% 14% 10% BOLNET 20% 15% 15% 5% 15% 30% REUNA 20% 10% 60% 3% 7% RdC 25% 45% 5% 15% 5% 5% Red Cetcol 30% 50% 5% 10% 3% 3% TEECOM-CO.N 1% 29% 22% 29% 14% 5% CENIAI 82% 3% 12% 3% ECUANEX 20% 5% 60% 15% RAIN 20% 60% 2% 10% 8% CNCnet 35% 15% 8% 15% 4% 23% LEDNET 50% 10% 10% 15% 15% RCP REDID 20% 0% 10% 0% 50% 20% RAU 70% 17% 4% 3% 6% REACCIUN 70% 1% 10% 10% 4% 5%
This questionnaire was administered between February and June 1995. These are the questions we asked, but they have been renumbered for exposition.
Network Contact person/point: Postal address: Telephone Number: Fax Number: e-mail address:
1. Description Operating since _____ Registered users _____ Active users _____ (connected at least once a month during the last 3 months) 2. Do you charge for services Yes  No  If Yes, specify charging methods (check all that apply) ____ Free to end users ____ Fixed charges to end users ____ Variable (usage-based) charge to end users ____ Free to end institutions (such as a university) ____ Fixed charged to institutions ____ Variable (usage-based) charge to institutions 3. What is the approximate annual budget for your network connection and/or information center? _________________USA$ 4. Of that budget, approximately what percent is used for _____ communication equipment _____ computer equipment _____ communication charges _____ technical staff _____ management staff _____ support staff _____ facilities _____ software _____ overhead _____ other ______________________ 5. Type(s) of connection(s) or gateway(s) out of your country _____ IP _____ UUCP _____ Fidonet _____ Bitnet _____ Other _______________________ 6. Communication links(s) out of your country (for each link) Speed ____ Leased or switched ____ Vendor _________ Where do you connect? ________ 7. Approximately, what percent of hosts on your network run ____ an FTP server ____ a WWW (http) server ____ a Gopher server ____ a list server ____ a news server ____ a dial-up bulletin board ____ a library catalogue 8. Approximately, what percent of the hosts in your network communicate via ____ IP ____ UUCP ____ Fido ____ Bitnet ____ Other ________________
1. Scope of the network (check all that apply) Academic  General Research  Government  NGO  Commercial  Special Interest Group  Specify________ Other  Specify________ 2. Approximately, what percent of your users would you estimate as being _____ University and research faculty staff _____ University students _____ Government employees _____ Commercial employees _____ NGO employees _____ Employees of International organizations, e.g., (UNESCO, PNUD, OAS, IADB, and World Bank) _____ Other ______________________ 3. Approximately, what percent of your users have ____ e-mail only ____ dial-in access to a command line account on a host ____ dial-in SLIP/PPP connectivity ____ full-time IP connectivity 4. Approximately percent of the users of your network have ____ Personal computers running DOS ____ Personal computers running Microsoft Windows ____ Personal computers running Unix ____ Personal computers running Mac OS ____ Unix workstations ____ Other workstations ____ Unix minicomputers ____ Other minicomputers ____ Unix mainframes ____ Other mainframes ____ Other ________________
1. Is there a help-desk or other central point for queries? Yes  No  Postal address Telephone number e-mail address 2. Is a general user guide available? Yes  No  Paper  Electronic  3. User group support activities Describe here activities to support user groups in using the network. 4. Training activities and workshops Describe here activities to train and inform user support staff. 5. Other activities Describe here any other activities on your network you consider relevant. Comments (plans, etc.)
Could you give some examples of users or applications which best illustrate the value of your network? Thanks for your cooperation!