Eugene F. Hastings II
Pittsburgh Supercomputing Center
4400 Fifth Avenue
Pittsburgh, PA 15213
During the last few years, thousands of schools and hundreds of thousands of students and teachers have begun to develop the Internet into a major educational resource. Within the community of educational researchers, the Internet is now an accepted standard. Network use is also growing rapidly throughout the larger educational community, but several factors have placed practical limits on this growth. This paper attempts to identify some of these limiting factors and suggest new technologies that can help schools surmount these obstacles.
Consider first some of the reasons that propel educational use of the Internet. These include the following:
Schools face possible obstacles in each of these areas. If a school cannot obtain adequate connectivity, its ability to overcome isolation will be limited. If it has no access to a server, its ability to publish materials online will be limited. If it has too few user devices for all students to participate in networking activities, the promise of equity will not be achieved. And if tools do not exist that enable students and teachers to use networks imaginatively and effectively to support their broader educational goals, the promise of educational reform will not be met. Overcoming each of these obstacles entails a technological challenge. In the following sections, we list these challenges and describe some mechanisms for meeting them.
Underlying our arguments is a common theme of shared infrastructure and shared support mechanisms. Networking is a means of building community, and a strong sense of community is often the key to building viable network-linked educational programs and resources and sustaining them over the long term.
The types of educational and community networks that we advocate differ in some essential ways from those developed for many commercial applications. Commercial networks emphasize privacy and security issues over cost and access considerations. For educational and community networks, cost and access are the primary concerns, and network architectures that neglect these factors are unlikely to be sustainable for very long.
Shared infrastructure and broad applicability bring other educational benefits. They create network environments that are conducive to interdisciplinary and collaborative activities and to research projects that explore an ever-expanding scope of resources. We believe that these sorts of activities and projects can contribute effectively to educational reform.
We find that the portions of networks with the highest development costs and the greatest need for cost reduction are precisely those with the fewest possibilities for shared infrastructure. To put the issue in colloquial terms, we must deal not so much with the problem of building the information superhighway as with building the information driveway that leads into the home or classroom. Once this part of network connectivity is established, there remains the task of using it effectively to access online material and produce materials for others to use.
How can we provide affordable access to the Internet for schools, community centers, libraries, and homes?
Two factors dramatically affect the performance and affordability of internetworking: the cost and bandwidth of the "local loop" from the user's site to the nearest network point of presence, and the existence of a complementary wide-area infrastructure to tie the various local loops together. These two factors are critical regardless of the specific details of the chosen technology or the market position of the service provider.
Schools and communities should choose their network technologies and services strategically, with an eye toward how their choices will support and leverage the overall networking infrastructure. Broad infrastructure needs are a much more important consideration in developing educational and community networks than in planning private networks or networks that serve a limited number of public or nonprofit organizations. Even though educational programs have a specific focus, they inspire other activities, and they rely on other organizations for many kinds of resources. Hence, it is important to consider the entire community as a unit when assessing alternatives for providing an economical infrastructure for school or community networking.
First, we take up the issue of providing the local loop. Several factors make it likely that the costs of providing connectivity will drop, perhaps dramatically, in the next few years. One strong influence is market competition to supply the local loop. Where there are multiple suppliers of connectivity to schools, homes, or offices, competitive forces will help reduce the cost of network services to these sites. Many communities are experiencing active competition in this area, although recent federal telecommunications legislation may allow the reestablishment of monopolies of local connectivity, with a corresponding decrease in options and likely increase in price. One must also take care to distinguish competition in the local loop from competition in limited market segments, such as voice or video services. While this latter form of competition could decrease the price of specific services, it may have no relevance to data services and internetworking. More typically, some service providers may engage in "cream-skimming"--seeking to exploit market segments with the highest profit margins, a practice which public utility commissions and local regulatory authorities should probably discourage.
Tariff policies also affect the viability of various options for Internet access. For example, the commonly available digital telephone service known as ISDN can provide economical connectivity at moderate bandwidths, as long as its tariffs are set at an affordable flat rate (with no time charges). When time charges are applied, this service becomes uneconomical for typical network interconnects, since it is desirable to keep the connection in place throughout the day.
Competition to provide local loops is likely to involve a number of new technologies. Telephone subscribers can employ new electronics--most notably, a technology known in telephonic circles as High-bit-rate Digital Subscriber Loop (HDSL)- -to wring vastly higher data rates out of older copper infrastructure. HDSL can provide data services at speeds in excess of one megabit per second and costs as low as one-tenth those of traditional connections. Public utility commissions will ultimately determine whether these economies result in savings for home, school, and office subscribers or in larger profits for telephone companies. This is a complex issue, since rates are determined not only by the intrinsic cost of any single service to the telephone company, but also by the fraction of equipment and maintenance left in the hands of the customer, the need to subsidize certain lifeline services, and the need to average rates across the many different technologies used to supply local loops for a given region.
In addition to HDSL, which provides a symmetrical service, there are also asymmetrical services, such as Asymmetric Digital Subscriber Loop (ADSL). Symmetrical services provide the same data rate into and out of a given site. Asymmetrical services typically provide higher data rates into a given site than out of it. While asymmetrical services may be acceptable for a residence, where the users are mostly consumers of information, they would not be appropriate for connections to local area networks at schools and libraries because they would limit the ability to publish information at these sites. And although some ADSL technologies have much higher bandwidths than HDSL, they provide significantly higher speeds only in the forward (outward) direction, and they cannot drive data for the distances that HDSL can. A new service known as SDSL promises to offer the bandwidth advantages of ADSL in a symmetrical mode.
Another promising development in the field of telephony is the deployment of fiber optics, either to drop points in the neighborhood or to the curbside of subscriber homes or offices. Fiber brings vastly higher bandwidth than copper and promises higher reliability over time. Once the capital costs of laying this new infrastructure are depreciated, suppliers can anticipate significant reductions in the costs of providing high- bandwidth data services, namely those with speeds in excess of 1 million bits per second. Such services would exhaust only a tiny fraction of the bandwidth available through a fiber infrastructure, which can also provide video services requiring the equivalent of tens of millions of bits per second.
Some telephone companies are contemplating a fiber optic/coaxial cable hybrid system, an architecture commonly used by newer cable television installations. These systems provide a particularly attractive architecture for data services and internetworking: they can offer services with peak bandwidths of 10 to 40 million bits per second or higher, at prices potentially competitive with those of present-day analog telephone service. Widespread deployment of such systems will help redefine the pricing structure for high-bandwidth data services, as prices drop as low as a few percent of current service offerings. With extensive deployment, the cable modems needed to deliver this service can be marketed at prices comparable to those of conventional modems, and the service itself can be sold for prices comparable to those of premium cable channels.
Where internal wiring is costly to install and where external connectivity is difficult to achieve through fixed links, wireless technologies offer an attractive option. With its ease of setup, wireless is particularly appealing in situations where costly and time-consuming building modifications are necessary to install internal wiring. Wireless links can also be used to provide pieces of a metropolitan area network infrastructure. But one must bear in mind that maintenance costs for wireless links are likely to be higher than those for fixed wiring and that the bandwidth available for wireless communications is extremely limited compared to the bandwidth of fixed copper or fiber infrastructure. This important point has been largely overlooked by groups advocating wireless technologies as a panacea for public network access. These technologies offer an excellent starting point for such access, but as usage grows, the bandwidth available through this medium will inevitably become saturated, and the majority of public Internet traffic will have to be transferred to traditional fixed communications links. Thus, there will have to be truly compelling reasons to deploy wireless communications between fixed points in a metropolitan area.
Wireless technologies can be used effectively, however, as a rapid enabler, a temporary connection that can serve a networking purpose while fixed wiring is being planned and installed. Upon completion of wiring, the wireless apparatus might be reused to help leverage another start-up site.
New technologies are also generating new means of tying together local loops to form a wide-area networking infrastructure. New telephone architectures offer the possibility of significantly reduced costs for data services and internetworking. The traditional telephone infrastructure connects lines from homes and offices to switches at the telephone company's central offices, forming circuits through which telephone calls are routed. This circuit-switched infrastructure is oriented toward point-to-point connections, with endpoints that may be very widely separated. Each connection ties up switch and transmission resources for the duration of the connection, regardless of how much information is being passed through the circuit.
Internet services are based upon a quite different architecture, one that involves a common "cloud" of connectivity through which individual data packets are routed to their final destinations. This packet-switched infrastructure offers a number of significant advantages for data transmission. Individual circuits need not be maintained, so there are lower costs in terms of switches and infrastructure relating to the transmission plant. Demands are made upon this infrastructure only during the actual transmission of data packets. Furthermore, thanks to the manner in which Internet protocols behave in the presence of traffic congestion, the system can be overloaded briefly with only a slowdown in the rate of data transmission as opposed to a failure to establish a desired connection.
This type of packet-switched infrastructure can be provided through a number of technologies commonly offered by the telephone industry and referred to collectively as "fast packet services." The optimal mix of technologies will depend upon local pricing and needs. More important than specific programmatic needs may be the overall strategies employed to provide connectivity to schools and community centers in a given area. The following points are relevant for schools and communities developing strategies for connectivity:
By working with local telephone suppliers and following these guidelines, communities can provide good connectivity at modest cost.
How can we help all students and teachers make the most effective use of their connectivity to pursue educational and social goals?
Connectivity alone does not create a networked environment of value to schools and communities. More important is the quality of information available over the network and the ability of schools and communities to use the network as their own publishing venue. The content challenge involves a set of technical issues, which we discuss in this section, and a number of social issues, which lie beyond the scope of this paper. These social issues relate to the evolution of the network community, the formation of networked collaborations, and the financing of network resources that exploit the full richness and depth of the medium. Technological factors can assist in these areas, but human factors ultimately will determine the network's success and relevance.
The architecture of low-cost online services is increasingly dominated by client/server mechanisms for delivering information. Determining the costs of such services requires an analysis of the costs of the client and server machines and the availability of software suitable to provide these services.
We consider first the question of client machines or access devices. Computers can be bought now for $1,000 to $2,000. This has been true for the last half-dozen years, and it is likely to remain true as long as the only forces governing these prices are traditional computer marketplace forces. It is not the price of the raw materials or the manufacturing cost that determines the retail price of personal computers. Rather, marketing strategies and advertising decisions set these prices. Producing a $200 user device capable of running a World Wide Web browser over standard telephone links is within the realm of possibility, provided that a large enough market exists for such a device. Many forces are converging to make it likely that such devices will be produced. These include the progress of game machines, which now typically contain powerful graphics computers as a standard component, the development of set-top boxes for cable television systems, and the need of schools for large numbers of cheap, networked computers.
With the rapid evolution of computer hardware, operating systems, and software, it is possible to pass still-serviceable hardware from business and government offices into the hands of schools and community groups, who could use it to gain basic Internet access at extremely low cost. The advent of Windows 95, with its significant new hardware demands, is beginning to drive many commercial users to invest in new hardware. But much of the remaindered hardware is still capable of running the older Windows operating system and supporting graphical interfaces to Internet resources. If commercial users donated remaindered hardware to schools and community centers, and if mechanisms to support this hardware were put in place, it might be possible to rapidly provide network access to all of the nation's classrooms.
There are many uncertainties associated with such a scheme. It assumes that the evolution of software for classroom application will be much slower than the evolution of commercial software, so that the Windows machines will be adequate for most educational applications for a number of years. And it assumes that new features in commercial software will not make the training that students receive with older machines and software irrelevant to the commercial workplace. Finally, it assumes that older machines can be supplied in quantities sufficient to provide school districts with a uniform and supportable computing environment. If these conditions can be met, the donation of older equipment could provide a significant boost to school and community networking efforts. This is only an interim approach and a partial solution for school and community networking, but it should allow communities to stretch significantly the dollars available for networking projects.
Whether the client machines are standard office computers or low-cost school and home computers, they will require affordable client software in order to access common information services. Under one widely used marketing model that works to the advantage of educational users, software vendors charge a fee for their server software but give away client software free of charge to all users or, more commonly, to educational users. The rationale for this pricing model is that it encourages access to online information, for which there may be an access fee. Vendors may also charge a fee for the construction of servers to groups that place information online for commercial access. Under this type of pricing arrangement, it is possible for educational users to obtain high quality World Wide Web and Geographical Information System browsers, among others. The practical consequence is that the client code is free to educational users but is well-maintained to insure the availability of solid code for all commercial clients.
Whether the market will continue to support this type of pricing structure depends largely on whether the majority of vendors adhere to standards for transporting information over the network. If large volumes of material are available through a small set of common standard protocols, there are likely to be low-cost or free products available to implement these standards. If one vendor obtains a near monopoly on the market, there is a danger of that vendor introducing proprietary protocols in its products and forcing educational users to purchase them at arbitrarily high prices. We would regard this as a classic monopoly situation that could be addressed under existing antitrust laws.
On the server side, there is a relatively broad range of options, consistent with the idea that one need support only a small set of network protocols to supply a very wide range of information services. With an increasing number of vendors supplying Internet servers for school use, we are fast approaching the possibility of a generic Internet server. Generic servers can be built on the most common hardware platforms with public domain operating systems at a fraction of the cost of servers commonly available just a year or two ago. Current networking projects and forthcoming network products are likely to produce enough examples of server setup and maintenance to make their operation practical in any school or school district. Further simplification of this task will occur as new software is developed to facilitate the management of these machines, both locally at the school site and remotely from a district office or contracted system management provider.
As described above, some vendors provide free client software but extract a fee, sometimes a very large one, for the server software. Fortunately, many vendors are adjusting server charges based on the features offered and the audience likely to use them. For example, systems that include advanced security features for financial transactions are likely to be sold at higher prices than systems meant for simpler information delivery systems. Here, too, adherence to a small set of network protocols makes it likely that low-cost or free implementation of information services will be available to educational and community networks.
It is important to realize that services like the World Wide Web have their limitations, too, and to consider mechanisms that will help overcome these limitations for educational and community users. For example, the Java servers developed by Sun Microsystems can surmount some of the limitations of the hypertext markup language that forms the basis of the World Wide Web. Java allows a server to pass new mini- applications to the client machine. Because these applications can be of greater complexity than standard Web pages and because they are not limited by the constraints of the hypertext transfer protocol that underlies all Web pages, the Java mechanism greatly expands the functionality of the Web environment. We expect similar developments to continue to enhance server software and improve environments for developing and presenting online multimedia materials. It should be noted that the act of passing executable code to a client machine carries with it a strong risk of introducing viruses or Trojan horses on the client machine, so this type of technology should be carefully monitored as it develops.
Advances are also being made in the area of Web editors, which simplify the task of creating resources to be placed online via the World Wide Web. These advances are reducing the costs of training students and teachers to create multimedia presentations for online distribution and increasing their ability to develop these resources as part of their standard classroom activities. A number of vendors have announced products in this area.
Two important tasks must be accomplished in order to extend Internet access to all students and teachers and enable them to incorporate Internet resources and techniques into their everyday educational activities. First, schools and homes must be provided with connectivity, and second, facilities must be developed that make it easy and inexpensive to create and access online resources. We believe that technologies exist to meet these goals--indeed, we believe that the ideals of universal connectivity for all the nation's classrooms and commonplace classroom use of the Internet can be realized by the end of the decade.
Of course, technical underpinnings are only part of the much larger issue of major educational reform. Human factors ultimately will determine the extent to which networking technology can assist the task of educational reform. But technology can significantly reduce the human effort that must go into educational reform and can ease the cost of the transitions associated with reforms. The real challenge is to use the technology effectively to unlock the imaginations and abilities of students and teachers across the nation.
[ Back to the Futures Page ]
Last modified January 12, 2000 (mhm).