Expanding International Email Connectivity--Another Look John C. Klensin and Randy Bush Copyright 1993, John C. Klensin and Randy Bush. May be freely copied and distributed only in its entirety. (Originally published in ConneXions June 30, 1996) Unlike the situation of thirty years ago, when almost all important scientific work occurred in Western countries, the scientific community is becoming increasingly international. Important work and areas of study occur all over the world. Collaborations and ability to access sources of data and other resources are increasingly important to scientific progress. In many fields, we see more and more inter-institutional collaborations on research and papers that draw on the strengths of each of these institutions. Exchanges of ideas and collaboration and review of proposals should not be limited to one country, or even to developed areas. Especially in such areas as the health and social sciences and in all of the various fields that study "global and environmental future" issues, participation of scientists in developing areas has become crucial. This is true whether the scientists themselves are indigenous to, or visiting in, those areas; indeed, as the community becomes more international, the distinction between the two is gradually becoming less clear. Local policies often reinforce the trend toward collaborations that require strong communications links. For example, as it becomes harder to move biological samples or historical artifacts across international boundaries, it makes increasing sense to do analyses and evaluations within the country of origin, then make the data available to both both domestic and remote parties. To the degree that relationships involving local and remote scientists and institutions become permanent and stable, the benefits tend to spread with improved research and more open and tolerant relationships with regulators and those not initially involved. Communication facilities based on computer networks, especially the low-level ones such as electronic mail, have become critical to these types of collaboration. The post is simply too slow to permit real interaction, and fax, while faster, does not lend itself well to group interaction, much less true collaboration on shared materials. The more the network connection infrastructure can be opened up, the more scientists can participate in international efforts on the basis of interests, skills, and knowledge, rather than based on where they happen to live or work. Similarly, data gathered or prepared in remote locations often must be transferred elsewhere for evaluation or analysis. The examples usually cited are climate or rainfall data, but similar issues arise with health and nutrition status, mortality and disease figures, and even certain types of economic statistics. Without computer networks, the options are to mail machine-readable media (often unacceptably slow and unreliable) to send the data via fax (typically requiring rekeying or the use of OCR techniques, which are not completely reliable), or long-distance international data calls at great expense. Many Latin American, Caribbean, and African universities tell of invited US scientists who would not come for sabbaticals or extended research visits because they would not have email access to their colleagues. The problem of being without network facilities is not merely one of keeping up correspondence, but of being isolated to the point of ineffectiveness if reliable email is not available. On the other hand, when networks are available, previously-unanticipated collaboration seems to come into being almost spontaneously. Again, the underlying causes seem to involve a latent demand that remains latent as long as joint work requires either the disruption of waiting for the post, the continual retyping of texts transmitted by post or fax, or the need to secure large budgets and approvals for extensive international travel. When computer networks are available, people quickly become comfortable with using them, the collaboration seems to happen often and quickly in many disciplines and work groups, and needs only a little bit of outside stimulus to get it started in others. These patterns in the scientific community have been paralleled by activities in various agencies and organizations concerned with development or assistance. Groups in one country need to communicate with those in another, and, when they can be made available, electronic mail and computer-assisted communications have significant advantages over other approaches. The demand from both communities has been present for some years as made evident by the "how do I access a network to stay in touch with colleagues at home while I'm in the field or to work with colleagues in remote locations from my home institution," inquiries which appear at very frequent intervals on popular network mailing lists and news groups. In what we might think of as first-generation low-end wide area networking, the response was "call home": connections from a terminal in the field to a centralized computer. The calls might be made with modems over international telephone connections or remote-connection PPSDN links (usually X.25), but the essential communication pattern was remote login to a centralized computer that hosted what was, in reality, a centralized email system. Gradually that type of arrangement became somewhat less decentralized: a single central computer that everyone dialed into was replaced by regional central computers with the same type of arrangements and some way of communicating among themselves. As computer costs dropped and modem technologies improved to permit data communications over low-quality lines at higher speeds and plausible costs, opportunities arose for true computer-computer connections, with people receiving and composing mail on their local systems, rather than trying to type while connected to remote locations. Of course, those opportunities have been taken advantage of, but often in a way that may inhibit positive long-term network developments. Organizations with a need to communicate with subsidiaries or collaborators can now establish single-purpose polled or dialout arrangements, typically using FidoNet or UUCP technology, that link the components of that organization, or the collaborators in a particular project, with each other. In any given situation, this may be reasonable and the arrangements can be established with a minimum of fuss. Unfortunately, there are also negative effects: o Participants in one activity tend to become isolated from non-participants and participants in other activities. o While inter-country communication may be facilitated, intra-country communication may be frustrated: either made impossible altogether or forced through very remote gateways. o While two separate private arrangements may be cost-effective, the third one rarely is and there are usually major advantages in not starting the second. If a single organization can afford one polled international mail exchange a day, better service for everyone can typically be arranged for the same costs by sharing resources and arranging multiple exchanges. If the user base is wide enough and can be expanded without regard to project boundaries, it has been shown time and time again that a minimal networking channel will quickly build a user base which soon fills that channel. Even in less-developed areas, the perception of value rapidly builds to the point that the user base then manages to find its own funding to continually increase the available bandwidth. o Private arrangements tend to satisfy, and then hide, demand. Resources being invested that could contribute to higher-quality connections for an area do not appear when needs or market surveys are performed. This is especially critical in situations in which the precondition for building a network infrastructure is the ability to demonstrate that the demand and users exist. This demand often does not arise until networks are actually seen to be in place and working, and people involved in private network arrangements might otherwise make major contributions to it. Looked at differently, private arrangements tend to be examples of classic cream-skimming behavior: the needs which can be most easily financed are met, making it more difficult to meet needs -- possibly more serious ones -- that are less readily financed. o Network arrangements set up for a particular project tend to collapse when that project ends, leaving people who had learned to depend on a certain level of communications and connectivity without it. The decision to install a private network -- or, more often, simply a private star-type mail polling arrangement -- often results from lack of understanding or consideration of long-term implications. Many of the people who are normally considered experts can make poor guesses and give poor advice if cost and technology tradeoffs are radically different than they are in areas with established telecommunications and network infrastructures. This situation prevails, almost by definition, in many less developed areas. Although the reasons may be different, it is also prevalent in many areas of Eastern Europe and other portions of the former Soviet area of influence. A very similar situation occurs in the US K-12 arena, where the cost of a single phone line can be a major administrative obstacle. Even when networks develop within an area, without significant impetus from "outside", the user community that drives the installations may turn out to be the wrong one in the long term. For example, the history of starting networks and network connections has predominantly rested in computer science and computer technology-oriented departments, businesses, and other data communities. There is some history of these communities constructing networks for their own use and then using various mechanisms - costs, perceived complexity, or lack of user support to then hoard the resource. Then, when computer user communities - scientists, educators, or the general public - for whom the computer is a tool for communication or computation (but not an interesting device in its own right) need access to networking technology, they often need to start over. At the same time, those user communities are much larger and, in many places, represent the largest potential user community. They may ultimately have access to greater political, cultural, and financial resources than computer technologists. At the same time, the community which is more computer-oriented can, and typically has, managed to establish communications when they see it as sufficiently important. While they may not understand optimal approaches for a given area, they usually have access to sufficient information to make something work. Other scientists and organizations have often had less success, since they do not have the technology readily available and may believe (or "know") many things that are untrue, such as the need to focus computer networks around mainframe systems or centralized remotely-accessed hosts or the requirement for very high-bandwidth circuits to do anything useful. These misperceptions are encouraged by publicity releases from vendors of high-end systems and by a media focus on the "next generation". These problems, and the behavior patterns that cause them, are not limited to international development or developing countries. For example, in the US K12 (kindergarten through 12th grade educational) community, few of the innovative teachers, the potential initial users who perceive the benefits in advance, have the political and economic power to affect their networking destiny. Network inertia and data hoarding are rife in the administrative infrastructure, often leading to either no connections at all or to restricted private single-function networks. Theories of technological trickle-down are routinely cited but regularly disproven in the data networking arena. All in all, it may have become a little bit too easy to set up a "network". Maximum effectiveness in network-building requires a different approach in which we phase out remote dialup arrangements in favor of "hosts" co-located with the users and where possible phase out -- and stop creating -- private arrangements in favor of shared connections and infrastructure. We need to improve our structure of information about who is interested and what is already operating in a particular area so that, if there is will to cooperate, every effort in a particular area reinforces every other effort and strengthens the links to the outside. This, in turn, leads efforts down the path toward regional backbones as the most effective mechanism for providing adequate bandwidth through, and out of, the countries and regions. Interested parties can best leverage their own needs into effective networks if they have information about other interests and activities. If existing sites are not inclined to cooperate with new ones, the best solution is to simply develop parallel infrastructure, gradually leaving them out. That, of course, requires the same databases, training, and information as would be the case if there were no existing connections to the area. But, if we fail to move in ways that consolidate efforts and lead to better communications and interconnections within areas and between projects and disciplines, we shall see increasing intercommunication and connection difficulties among people and groups who "have email" or are "connected to networks". ------------------------------------ JOHN KLENSIN holds an S.B. and Ph.D. from MIT. He is director of the INFOODS Secretariat for the United Nations University and was until recently Principal Research Scientist at MIT. He has worked on or led major projects in statistical and scientific database management, interchange of very complex data, information location and retrieval with uncertain classification, data analysis and modelling, and the impact and influence of communications. He has tried to use computer networks to enable applications and non-expert users since the early days of the ARPANet, and has occasionally succeeded. He is a member of the Internet Engineering Task Force and chaired the recent working group on extensions to the SMTP protocol. He is also a member of ACM and immediate past chair of its Standards Committee, IEEE, the American Statistical Association, and the International Association for Statistical Computing. He can be reached as Klensin@INFOODS.UNU.EDU. RANDY BUSH is a compiler netware, and tools hacker, and too often a software engineering manager. Residing in Portland Oregon US, he is currently a software architect at Olsen and Assoc., Zurich. He has been a user and occasional implementor of networking for a few decades, and is a member of the Modula-2 language committees and other lost causes. He has been involved in in integration of appropriate networking technology in the developing world for over four years, using FidoNet, UUCP, and TCP/IP. His email address is firstname.lastname@example.org. Dr. Klensin and Mr. Bush are the principals in the Network Startup Resource Center, an effort to provide technical assistance to people developing computer networks, especially low-end ones in developing areas. That effort, and the development of this paper, are supported in part by NSF Grant No. NCR-9216064.