As of 22 April 2009 this website is 'frozen' in time — see the current IFLA websites
This old website and all of its content will stay on as archive –
UDT Series on Data Communication Technologies and Standards for Libraries
Packet Radio: Applications for Libraries in Developing Countries (1993)
3. OVERVIEW OF PACKET RADIO TECHNOLOGYPacket radio technology is the application of packet switching techniques to radio. Together, packet switching and radio provide an efficient mechanism to support wireless computer communication over a wide geographic area. Recent technological advances have made packet radio communications more effective and affordable, and thus a candidate for use in developing countries.
The subject of wireless data communications is complex and highly technical—involving radio communications, spread spectrum techniques, packet switching, multiplexing, and communications protocols—and there are many works that present thorough, detailed discussions of packet radio technology (e.g., Lynch and Brownrigg, 1987). Rather than repeat those efforts here, the present chapter provides only a general overview of the concepts underlying wireless data communications and packet radio technology.
3.1 Underlying concepts of packet radio
3.1.1 Wireless communicationsTraditional data communication is achieved by use of a fluctuating electrical signal over a wire connecting two communicating systems, a technology that has ultimately evolved from nineteenth-century telegraphy and telephony. More recent technologies have replaced copper wire and the electrical signal with thin glass fibre and incredibly short flashes of light. Both these kinds of wire-based, or "wired" communication have evolved rapidly in the past two decades so that now it is possible to send hundreds of millions of bits of information across a wire every second.
Wireless transmission, on the other hand, is achieved by using fluctuations in one of several levels of electromagnetic radiation to carry signals rather than electric currents in wire.
The telephone system is probably the most familiar example of point-to-point communication. A telephone call takes place only between a calling and a called party—it connects two specific points in the telephone network. A single telephone call, however, can involve both wired and wireless communication; the call might be routed via a satellite, in which case the transmission is a radio signal between a ground station, the satellite and another ground station; the call might be sent long distances via microwave links; for some parts of its trip the call will be carried on optical fibre and it finally travels over copper wire from the subscriber's local office to the telephone. All of these point-to-point communication modes are used with computer data communications.
Some radio signals, particularly at high frequencies, as well as microwave and light, require that there be a clear, unobstructed path from transmitter to receiver. In other words, the receiver must "see" the transmitter. Low frequency radio waves, on the other hand, penetrate most materials, and do not constrain receivers to be within sight of a transmitter. Although the use of line-of-sight paths would greatly simplify radio-based communication network operations, such a constraint would seriously limit the flexibility of a network in terms of equipment siting and use in a mobile environment. In practice, radio networks have to be designed to support non-line-of-sight broadcasting.
The design of the network must strike a balance between powerful transmitters that have a large range and causing interference with other, more remote, users of the same frequency. One means of doing this is to use low-power transmitters with many relay stations. This technique, however, makes the installation of the network more expensive, and requires the use of strategies to ensure that data is transmitted correctly from the sender to its destination via one or more intermediaries.
noise. The atmosphere is full of a low level of RF radiation. Sunspots, distant galaxies, electric motors, vacuum tubes (TV picture tubes), ignition sparks in automobile engines, all emit RF radiation at varying frequencies and strengths. This means that the perfectly regular signal emitted by the transmitter is immediately subject to interference and loss of regularity. This corruption shows up in voice radio as a continuous and varying level hiss, or noise. The level of noise will vary with the location of the receiver. A receiver located near a strong noise source, such as a busy highway, will be subject to levels sufficiently strong to interfere with the reception of data.
multi-path interference. Just as light waves are reflected from surfaces such as a mirror, RF radiation is reflected from many kinds of surfaces. The degree and angle of reflection varies with the surface, thus, when radiation is reflected, a single signal can take multiple paths to reach a receiver. Because some of these paths will be longer than others and RF radiation travels at a finite speed, they will arrive at the receiver at slightly different times. These tiny differences cause interference effects in the signal that are manifested as distortion and temporary fading of the signal. Such distortion has the effect of limiting the data transfer rates that the network can support. At higher frequencies, multi-path interference and noise are generally less of a problem, though the tradeoff is that signal attenuation is increased because of atmospheric absorption.
Signalling in the ground radio environment in relation to the above problems is too complex to treat in any substantive manner in this report beyond what has been briefly outlined. For a detailed discussion, see Kahn and colleagues (1978). Kahn suggests that packet radio systems should use radio frequencies in the upper VHF band (closer to 300 MHz), in the ultra-high-frequency (UHF) band from 300 MHz to 3 GHz, and in the lower portion of the super-high-frequency (SHF) band from 3 GHz to 30 GHz. Kahn also discusses the characteristics and implications for data communication of propagation path loss, multi-path interference, and human-made noise at these frequencies.
Essentially, spread spectrum techniques:
In general, when spread spectrum techniques are employed, the carrier signal is spread in a pseudo-random, but predictable, manner and the information signal is left unchanged. Both the transmitter and receiver are equipped with an identical pseudo-random carrier signal generator, thus the receiver can reconstruct (i.e, "de-spread") the transmitted information signal.
Spread spectrum, however, also poses special design problems. First of all, special radio transceivers are required that can transmit and receive on a very wide frequency band, or that can switch frequencies accurately many times per second. Secondly, transmitters and receivers must remain highly synchronized (Lynch and Brownrigg, 1986).
Spread spectrum techniques are now being used for wireless LANs, primarily because the use of spread spectrum allows sufficiently low-powered transmitters to be used that regulatory agency licensing is not required (Davidovici, 1990).
A more technical discussion of spread spectrum techniques can also be found in Kahn and colleagues (1978).
One means of improving the efficiency of use of a channel is to have several conversations going on simultaneously, a process known as multiplexing. In one form of multiplexing, each pair of communicating systems is allotted certain time slots when communication is permitted; other time slots are allocated to other users. This is known as time division multiplexing. Even in time division multiplexing, however, many time slots will be empty because there is no current communication and, in other circumstances, there will not be sufficient time slots available to convey all the information waiting to be sent.
Packet switching was devised to enable dynamic multiplexing of many communication sessions on a single channel. It is based on the partitioning of long computer messages into small, independently transportable parcels called packets, and the interleaving or multiplexing of packets from several computer sources onto each network channel.
A packet contains data of variable length, typically up to a maximum of a few thousand bits. It also contains all the addressing and control information necessary to enable the packet to be directed from its source through any intermediate network nodes to its final destination, and also for any transmission errors to be detected and possibly for some types of error to be corrected.
Packet switching is a store-and-forward technology. In a packet switching network, it seldom happens that there is a direct connection between the sender of a packet and its intended recipient. Instead, the packet is relayed via one or more intermediaries before reaching its destination. Wire-based packet switching does, however, require a physical connection between the sender of a packet and its recipient, whether that recipient is the ultimate destination, or an intermediate relay. Each of these intermediaries will retain the packet for a brief period to ensure that: 1) the previous packet has been received at the next station, or hop; 2) to send an acknowledgement of receipt of the packet back to the sender; or 3) to determine the most efficient connection for sending the next packet. This delay is normally very small, in the order of fractions of a second, and some packet switching protocols implement techniques to reduce this delay even further. The primary international standard for packet switching is the CCITT X.25 standard, first recommended in 1976.
Packet radio networks have a unique set of features that set them apart from other types of packet switching networks and make their design more complicated:
a broadcast medium is used to communicate with a specific receiver, or very few receivers. Broadcast media are convenient for applications such as voice radio or cable television because the signal is intended for use by every receiver. In data communications, however, the signal is intended for a specific receiver (or a small group of receivers). The fact that the signal is broadcast to a specific receiver imposes additional requirements to ensure effective communication. The use of a broadcast medium in this manner requires that two conventions have to be observed by transmitters and receivers: 1) the transmitter must know the address of, and explicitly direct the packet to, a single recipient; and 2) all receivers have to agree to process only those packets intended for them. Packets not addressed to them should be retransmitted or discarded according to the rules of the particular protocol in use. It should be noted that the broadcast nature of packet radio makes it difficult to protect the security of the data without encryption.
packet radio uses relatively low-powered transmitters. This means that not every network node can receive the radio transmissions from every other node. Thus, packets must be relayed by intermediate nodes as is done in other types of packet switching networks. This necessitates strategies for ensuring that packets are correctly routed from the sender through one or more intermediaries to the final recipient.
packet radio transceivers cannot transmit and receive simultaneously. Unlike wired communication, in which the input line is normally different from the output line allowing simultaneous two-way communication, a packet radio transceiver cannot simultaneously receive and transmit data. Instead, a complete packet must be received before a packet can be transmitted. This may have a considerable impact on the network's throughput—or the speed at which data can be passed through the network.
all the nodes are built identically and serve as both data entry devices and as relays. Each node includes a radio transmitter and receiver in addition to a packet switching computer.
Because there are no physical connections between network transceivers (i.e. no wires), it is easy to move the transceivers. Thus, many packet radio networks must be designed to accommodate mobile transceivers, and particularly to accommodate a changing set of network nodes as receivers move into and out of range of the network. This feature of packet radio networks, however, is not of particular concern to library applications in which nodes are not mobile, with the exception of bookmobiles.
The following sections summarize the basic capability and service requirements of a packet radio network that are of particular interest to the library network environment (extracted from a set of general packet radio network capabilities and services identified in Kahn and colleagues, 1978). The capabilities and services are stated as requirements rather than inherent features of a packet radio network. Because it is a developing technology, not all packet radio networks provide the same features. However, all packet radio networks are based upon the same underlying concepts and principles, therefore most of the basic requirements will, in general, be satisfied.
The basic internal operation of the network should be transparent to the user. All user data presented to the network should be delivered to its destination without modification in any way. Only the data to be delivered, and the necessary control and addressing information should be required of the user as input. All other aspects of routing, reliable delivery, protocols, and network operation should be handled by the network itself.
All valid traffic originators within the network should be provided with connectivity with all other valid receivers subject only to the overall reliability and performance of the network. The network need not have prior knowledge of which users may wish to connect to other users or resources in the network.
The packet radio network should be capable of internetworking in such a way that a user providing a packet with an address in another network can expect the packet to be routed to a gateway to the other network or an intermediate network for forwarding. Similarly, arriving internet packets should be routed to the appropriate packet radio network user. This capability is particularly useful once all the resources in a particular library packet radio network are exhausted at which point a wider resource base can be accessed through internet connection.
Radio frequency characteristics of the packet radio system should allow coexistence with existing users of a chosen frequency band. This would provide a greater degree of frequency spectrum sharing, particularly among similar systems, and may facilitate the introduction of the technology in new geographic areas. As an example of coexistence, the RF characteristics of signals of the familiar AM radio band allow several radio stations to coexist without interfering with one another.
Throughput and low delay
The throughput and delay of the packet radio network should be sufficient to provide real-time interactive services, and to accommodate efficient data transfers. The specific requirements will largely depend on the requirements of the particular applications running on the network. The important aspect is that the greater the number of packet hops through the network, the greater the end-to-end delay. In addition, the specific multiple access protocol implemented in the network greatly affects throughput.
For wire-based communications systems, bit error rates from 1 in 100,000 to 1 in 1,000,000,000 can be achieved. With such bit error rates, the well-known ARQ (automatic repeat request) mechanisms (i.e., transmission, followed by acknowledgement or negative acknowledgement, possibly followed by re-transmission) are sufficient to ensure high efficiency data transmission.
Packet radio networks may have higher bit error rates (on the order of 1 in 100 to 1 in 10,000) due to the more hostile signalling environment. With such error rates, pure ARQ mechanisms may be insufficient due to frequent re-transmissions. In such cases, to maintain satisfactory throughput and delay more sophisticated technology such as forward error correction (FEC) must be employed.
In forward error correction, additional bits (typically 3 for every 8-bit byte) are calculated and transmitted to enable the receiver not only to detect but also to correct certain kinds of transmission error. The use of FEC requires that more data be transmitted, and imposes a degree of processing overhead at both the transmitter and the receiver, but their use enables standard acknowledgements to be used on the network, leading to acceptable throughput rates.
The network should support efficient packet routing between any pair of users on the network. Another useful capability of packet radio networks is the ability to "broadcast" a packet to a subset or to all users on the network. Although there may not be a requirement for this capability in library applications, such a capability is very useful for other applications, such as dispatching.
There are a number of other capabilities that are typically required for military packet radio network applications and are not required for library applications. These include area coverage, mobility, rapid and convenient deployment, addressing options, and resistance to jamming, spoofing, detection, and direction finding.
Wireless LANs may be based on one of several technologies. Most common at present is the use of infra-red light for signalling, using the same principle as a television remote control. The use of infra-red light requires a continuous line of sight from the transmitter to a receiver and is achieved by use of a reflecting ceiling. The architecture of these networks is similar to that of a single multiple access channel with carrier sensing and collision detection, much like Ethernet. Such infra-red networks can only connect computers located in the same room.
Mobile data services
Mobile data services, such as Ardis, Mobitex and CoveragePLUS in the U.S. (see Mello and Wayner, 1993; Brodsky, 1990), are nationwide data radio services where subscribers are charged fees to transfer data over a mobile data network. The service providers target specific markets, such as field service organizations, vehicle tracking and trucking, and offer specific service areas (i.e., states and cities) and coverage types (e.g., in-building, on-the-street, interstate highways). The RF (radio frequency) link protocols are typically proprietary, thus requiring proprietary radio equipment to interface to the radio network. These systems typically use separate frequencies for transmitting and receiving, and require that all data is routed via a central switching site.
Another method of data communication is to use the cellular telephone systems that have spread across most industrialized countries. Although these systems are based largely on analogue rather than digital technology, they have the advantage of an installed base of cellular transmitters, and relatively inexpensive cellular radio modems (Mello and Wayner, 1993). Data communication through the cellular network is much like using the regular phone system. All that is needed is a telephone—in this case a cellular telephone—and one of the many commercial cellular modems and adapters that plug into it.
The cellular modems are more sophisticated, however, than standard wire-based modems because of the special characteristics of the cellular system. Specifically, the cellular system works by covering a geographic area with a blanket of contiguous "cells" about eight miles in diameter—the radio range of a transmission tower. As a caller moves along, the signal must be passed off from cell to cell as the caller moves into and out of range of the transmission towers. When the call is passed off, however, there is a break in the signal that can last up to a second caused by changing transmission towers and frequencies (adjacent cells do not use the same frequency). Cellular modems are designed to compensate for this environment by using techniques such as error correction, data compression, as well as variations in transmission speed and packet sizes (Mello and Wayner, 1993).
An important method of wireless communication is the communications satellite. The position of satellites in orbits high above the earth allow them to relay information over vast distances as measured on the ground. Satellites relay voice and data by transmitting signals received from the ground (the uplink) back to another location on the ground (the downlink). Two different signals are used for the uplink and downlink signals to avoid interference at the satellite (de Sola Pool, 1990).
Satellites can vary along a number of dimensions: power, orbital path (e.g., geosynchronous/nonsynchronous), frequencies utilized, and width and direction of the beam. There is a correlation between the frequency the satellite uses and the size of ground station needed to receive a signal; signals from sophisticated, satellites using high frequencies can be received by small, low-cost ground stations, while satellites using lower frequencies require larger, more expensive systems on the ground (de Sola Pool, 1990).
VSATs. A technology that has great potential for use in developing countries are Very Small Aperture Terminals (VSATs). VSATs are small satellite earth stations that receive signals from high power Ku-band satellites and have antennae measuring less than two meters. They are low cost, simple to control, flexible, and adaptable.
New developments—such as microwave monolithic integrated circuits, very large scale integration chips, and smaller antennae—have reduced the cost of VSAT technology. In 1988, for example, a Ku-band earth station was cited as costing less than $1000 (Communications News, 1988). Once a VSAT system has been implemented, the cost of adding additional remote sites is nominal compared with other terrestrial-based communications technologies (Stalberg, 1990). VSATs are useful in providing communications systems in rural areas where no facilities exist and are "currently the fastest-growing application of communications satellites" (Stalberg, 1990:94).
LEOs. LEOs, or low-earth-orbit satellites, are non-geosynchronous and orbit an average of 850 Km from the earth. Their low orbit allows the use of low-powered transceivers and antennae. The cost of building and launching LEOs is also economical, in comparison with larger geostationary telecommunications satellites (1/400 of the cost) (Bennett, 1992). Communications with single LEOs, as with all non-geosynchronous satellites, is non-real time. Volunteer's in Technical Assistance's satellite, for example, can communicate with a specific ground station for 12 minutes, twice a day, allowing the transfer of 50 pages of data (Marek, 1992). However, grids of LEOs can be used to create communications systems that can function in a real time mode. Examples include Motorola's Iridium, TRW's Odyssey, and Loral/Qualcomm's Globalstar (Frieden, 1993).
The radio transceiver—the familiar two-way radio—is the device that provides connectivity to other radios in a network. It transmits the signal produced by the Terminal Node Controller (TNC) via RF radiation. One advantage of using transceivers is that existing two-way radio stations can be used in a packet radio network. However, Garriott (1991b) points out that newer radio technology is preferred because it more easily accommodates packet communications.
The Terminal Node Controller
The device used for encoding digital data into radio signals—the interface between the radio and the computer—is the Terminal Node Controller (TNC), also know as a packet controller, a packet assembler/disassember or a frame assembler/disassember (Mayo, 1989). It breaks the digital information from the computer into packets and transmits them through the radio via audio tones. At the receiving end, the TNC converts the audio tones back into packets and sends this reconstructed digital information on to the computer. Thus, it acts much like a standard modem that enables computers to communicate through standard analogue telephone lines by converting digital information to analogue information, and back again. The TNC also performs error-checking, as well as many other functions (Garriott, 1991b).
Commercially Available Packet Radio Modems
Commercially available radio modems combine the radio and terminal controllers in one device. Commercial radio modems incorporate sophisticated digital signal processing techniques, such as forward error correction, to achieve higher data rates than that of TNC-based packet radios. Such techniques, however, are typically proprietary. Tetherless Access, Ltd, for example, is a company based in California that manufactures radio modems.
The TNC or radio modem connects directly to a microcomputer. Although any microcomputer can be used, laptop computers are considered to be the best choice for applications in developing countries because they are highly portable, robust, and have low power requirements (Cannata, 1991b).
Where electricity is unreliable or absent, the addition of solar panels can provide needed power. A bank of batteries can help modulate the availability of electricity during nighttime or periods of low sunlight.
There are seven layers in the OSI Reference Model:
Data link layer. The second layer, the data link layer, accomplishes the error-free exchange of data between two directly connected systems.
Network layer. While the first two layers handle exchange of data between directly connected, adjacent systems, the third layer, the network layer, provides services that enable systems anywhere to exchange data.
Transport layer. The transport layer, layer 4, is the first in which data is transparently exchanged between end systems; all the lower layers may involve the processing of data by intermediate systems to ensure, among other things, proper routing and error correction.
Session layer. The session layer adds structure to the connection provided by the transport layer.
Presentation layer. The presentation layer ensures that information is transferred in a form mutually intelligible to different, possibly incompatible computer systems.
Application layer. The application layer is the highest, and most complex, layer of the OSI model. It is important to note that the application layer does not specify what applications are to be available to a user of a computer system, nor how any particular implementation should appear to a user. Rather, the application layer specifies what communications services are to be available to a computer system for a number of specific purposes.
126.96.36.199 Packet radio protocol issuesBecause of the hostile broadcast radio environment within which packet radio must operate, the protocol issues that must be considered are almost entirely at the lower 3 layers—the physical, data link and network layers. Problems at higher layers are the same as in other networks, except for the fact that packet radio protocols impose some limitations on data transfer rates, which, for example, may make it impractical to transfer large files in real time. Brownrigg, Lynch, and Pepper (1984) present a full discussion of protocol issues, which is summarized here.
Physical Layer. The physical layer enables and defines the procedures for transfer of data over a wireless broadcast channel. Among the issues which need to be considered in defining the physical layer implementation of a packet radio network are: choice of frequencies, modulation techniques, transmitter power, antenna configurations.
There are as yet no well-defined standards for packet radio at this layer. In addition, choice of frequency is highly constrained by government regulation and intense and increasing competition for the finite set of radio frequencies.
Choices of protocols made at this layer will influence choices to be made later on. For example, spread spectrum encoding techniques offer greater reliability of data because they are more resistant to problems such as interference and fading. On the other hand, spread spectrum is considerably more complex and more expensive to implement.
Data Link Layer. The data link layer ensures error-free transfer of data between two points. Issues to be considered in designing a packet radio network are: channel access, frame (packet) size, acknowledgements, error detection and correction. This is where multiple access channel protocols must be implemented (see below). Again, there are as yet no-well defined standards for packet radio at this layer, although the problems are similar to those faced by other broadcast networks such as Ethernet.
Network Layer. The network layer ensures that data reaches its intended destination across a network via one or more intermediate links. Issues to be considered in designing a packet radio network are: addressing, routing, and internetworking.
Routing within the network will be an issue only if the network is large enough to require repeaters. In a smaller network, every packet transmitted is received by every receiver, therefore no routing protocols are required. In a larger network, however, not every receiver will receive every transmission. Therefore, there must be procedures for routing a transmission from a sender via one or more intermediates to its final destination.
Another issue to be considered is that of internetworking; that is, providing connectivity from packet radio networks to other kinds of networks, such as X.25 public data networks. This requires at least one station in a packet radio network to act as the internetworking gateway, and complicates questions of addressing and routing.
CSMA/CD, or CSMA with collision detection, is an improvement over CSMA. In CSMA/CD, two stations that have begun to transmit data simultaneously during an idle period abort their transmissions immediately upon detecting a collision, instead of finishing the transmission of their frames. This immediate cessation of transmission conserves time and bandwidth (Tanenbaum, 1988).
In 1-hop networks the source and destination nodes are directly connected to one another, however, in n-hop networks the stations are not directly connected and packets must be relayed by intermediate nodes. In such a network, the throughput efficiency of CSMA protocols can degrade to the level of ALOHA protocols. This is due to the problem of "hidden stations" where the transmissions of two or more stations can collide at an intermediate station in between the two transmitting nodes. In this case, the two transmitting nodes are out of the receiving range of one another, therefore they cannot detect whether the other is transmitting. Thus, the benefits of CSMA collision avoidance cannot be applied.
Busy Tone Multiple Access (BTMA). To improve the throughput of n-hop networks, the conventional protocols of 1-hop networks can be modified to take into account the special physical characteristics of an n-hop packet radio network. Busy Tone Multiple Access (BTMA) can be applied in which information about hidden stations is broadcast (Tobagi and Kleinrock, 1975). In this scheme, a busy tone is sent on a channel separate from the data channel to indicate neighbouring stations not to send data (i.e., not to disturb an ongoing transmission).
Conventional frequency division multiplexing (FDM) or time division multiplexing (TDM) techniques can be applied which split the channel bandwidth into frequency or time transmission slots, thus enabling parallel communication of different stations.
Code Division Multiple Access (CDMA). Another possible multiplexing technique is CDMA (Code Division Multiple Access) which is the use of spread spectrum techniques to split up the channel into subchannels by associating a special code with each subchannel (Pursley, 1987).
A package available in the public domain, called KA9Q, has been developed specifically for this purpose. KA9Q uses the Internet Protocol (IP) over AX.25, with the internet address resolution protocol (ARP) for address mapping. Higher level TCP/IP protocols (i.e., TELNET, FTP and SMTP) are thus available for use.
TCP/IP was developed initially for the U.S. Defense Advanced Research Projects Agency (DARPA) from the mid 1970's (Comer, 1988). In part, its widespread use stems from the fact that TCP/IP was incorporated into a version of the UNIX operating system and, by this means, became readily available to the academic and scientific communities who were the largest users of UNIX systems. Further, products implementing TCP/IP support communications among more hosts and different types of data links than any other protocol suite. The types of hosts connected through TCP/IP range from Cray supercomputers to IBM PCs, and Apple Macintoshes. TCP/IP can be carried over Ethernet, optical fibre links, DECNET, X.25 networks, satellite links, and hi-speed telephone lines. Thus, it has become the basis of the Internet, the large, worldwide network of networks.
The two main protocols in the lower layers of TCP/IP suite are:
Transmission Control Protocol (TCP). TCP exists at layer 4 (Transport) of the OSI Reference Model and establishes "virtual circuit" using the connectionless packets of the IP. It discards duplicate IP datagrams, asks for re-transmissions of lost datagrams, and reassembles the datagrams into the proper sequence.
TCP/IP application protocols—those protocols that define the actual applications that users see—include the File Transfer Protocol (FTP), Simple Mail Transfer Protocol (SMTP), and the TELNET remote login protocol.
Email Electronic mail is the most common of network services and, on TCP/IP networks, is defined by the Simple Mail Transfer Protocol (SMTP). This format uses a simple mail format, with the message represented entirely as lines of text, using 7-bit US-ASCII encoding. The basic format uses a set of headers, containing data such as sender's name, recipients' name, subject, and date, and a body which contains the text of the message.
Because the format of the mail message is entirely text-based, transfer of binary data is not possible, nor is support available for any character set other than the basic English alphabet. Thus, the protocol is awkward to use with foreign languages, and impossible for non-Roman scripts such as Japanese or Cyrillic. Neither is it possible to enclose binary data as all or part of a message, such as images, documents embedded with word processing code, or spreadsheet files. A new Internet email standard, called MIME (Multi-purpose Internet Mail Extensions), is design to overcome this limitation by allowing binary information to be included in the message (see Borenstein and Freed, 1992).
TELNET is the TCP/IP remote login protocol. It allows a user to connect to a remote computer and interact with it as though his or her local computer were a terminal of that remote machine. The protocol defines a Network Virtual Terminal to provide a uniform format for transmission of data over a network. The user invokes a TELNET client application which translates between the format of the data as output by his terminal and the format as transmitted by a Network Virtual Terminal, while the remote system has a TELNET server application which translates between the format of the Network Virtual Terminal data and the format required by the host application.
The format used for transmission over the network is very simple, consisting of lines of US-ASCII text, including control characters and escape sequences. The protocol therefore does little more than pass characters from a terminal to the remote application and back again. This raises potential problems of interoperability between the application and the terminal. For example, the application might be designed to be used by dedicated, "hard-wired" terminals and will therefore not provide mechanisms for starting and stopping communication sessions gracefully. The application may assume the use of a single terminal type and be highly tailored to that, in terms of both screen manipulation and messaging.
FTP File transfer, as the name implies, involves sending files from one computer to another. Unlike electronic mail, which relays data from one computer to another and stores the message until the user can read it, file transfer establishes a direct interactive connection between computers to send a file. On TCP/IP networks, file transfer is known as FTP, for File Transfer Protocol. In operation, a user logs into a remote machine, examines the contents of file directories, and uploads or downloads copies of files. The user breaks the connection once he or she has finished. FTP can be used to transfer bitmapped images, documents formatted with word processors, executable programmes, spreadsheets, as well as plain ASCII text.
One of the more common uses of FTP is the convention for "anonymous FTP", which allows a user to retrieve selected files from a remote system without prior authorization. This allows for rapid and informal publication of information across a network, without needing to mail potentially long documents to many users. Both text and program files are frequently distributed in this way.
As a general rule, the higher the RF data rate, the higher the cost of the radio equipment. Since cost is a primary requirement for use in developing countries, appropriate equipment will necessarily be at the low-speed end of the spectrum. There are, however, arguably many different "acceptable" throughput rates, dependent upon the context of use. For example, acceptable throughput in a production-oriented industrialized world library for document delivery using images would probably be very high. In contrast, in developing countries where the alternative may be ground mail transfer taking several months, very low throughput rates may be acceptable.
performance requirements such as the required throughput and geographic scope of the system.
specific applications and their effect upon the design of the network (e.g., packet radio systems have been designed for symmetric data rates, whereas many library applications are asymmetric; Brownrigg, Lynch, and Pepper, 1984).
network topography such as whether repeaters are needed and their number, density, and location.
the frequency and RF bandwidth. A suitable radio frequency must be chosen that provides necessary bandwidth and area coverage (dependant upon constraints imposed by RF communication regulatory commissions). The choice of frequency and the design of the network topology is not a simple one and is dependent on several interrelated issues (see Lynch and Brownrigg, 1986).
hardware such as transmitter type and power; data rates and capabilities of the radio modem; and antenna design (including height and directionality).
signaling, encoding, and modulation techniques.
choice of protocols such standard or slotted ALOHA, or CSMA.
available funds for network development. This factor is, of course, a prime consideration in developing countries.
Clearly, each packet radio network must be designed to satisfy the unique set of requirements needed to carry out a given purpose in a given environment. The "right blend" of elements, such as "transmission frequencies, modulation techniques, transmitter power, and antenna configurations" must be found (Brownrigg, Lynch and Pepper, 1984:234). Thus, each packet radio network will be a unique solution to a specific connectivity problem.
Example packet radio models
Because of the very wide range of situations where packet radio could be applied in developing country libraries, it would be impossible to provide a blueprint or a "how to" guide for constructing packet radio networks. Nonetheless, it would be useful to highlight several "models" of packet radio networks to provide a context for discussion of specific categories of library network applications in the next chapter. These categories of applications, for example, include access to centralized databases, connectivity among libraries in a given region, and connections to large national and international networks.
Four basic models of potential packet radio networks are presented below. The specific technical details of their design are not dealt with, rather, they are presented in a general way as broad examples of packet radio networks that may be useful. For those wishing more detail about specific design choices for packet radio network and the interrelationships among them, please see Leiner and colleagues (1987) or Lynch and Brownrigg (1987).
Most communication in this 1-hop network would be between the central site and the remote stations, with little communication from remote station to remote station. The central site would act as an intermediate system that relays messages between remote sites. Thus, there would be no need for repeaters, routing processes, and message relaying in the stations of this network. The network will support some messaging that is station-to-station, however, this would be routed through the central switch.
The 1-hop topology requires all stations to have sufficient power to reach the central site. The choice of frequency is thus dependent upon the distance separating the remote and central stations. For long-distance transmission (i.e. over-the-horizon transmission), HF (shortwave) band frequencies may be required. The transmitter technology for this band is cheap and widely available. However, the HF band provides a very unreliable data channel (e.g., noise, atmospheric disturbance, scatter, and multi-path interference), and a Figure 2 packet radio system will need more complex software to use it effectively. Transmission at higher frequencies would be possible, but it might require tall and/or directional antennae (e.g., antennae mounted on roof-tops).
One constraint upon choice of frequency is that a suitable frequency must be available that complies with licensing and regulatory restrictions.
Central packet radio station
The central site in this network is the most complex component. It requires a computer system that can, for example, process incoming messages from multiple sources, transmit to multiple destinations, and retrieve and manipulate data corresponding to multiple requests. The central site also requires software for routing packets at the network layer for packets addressed from one remote site to another remote site.
A small computer based on a microprocessor such as the Intel 80386 and using a multi-tasking operating system, such as UNIX, could suffice. Much depends, however, on the level of traffic and what applications are offered at the central site.
The remote stations in this network would be much simpler since they only need to support a single communication session with a single destination at any given time. Similarly, they will only need to support a single application process at any one time (i.e., they need not be capable of multi-tasking). This would allow relatively inexpensive computers to be used, such as laptop computers based on the Intel 80286 microprocessor. The components of the remote stations would consist of a TNC and receiver, or a radio modem, to handle the physical and data link layers. The workstation could be used both for local processing needs, such as word processing, accessing a maintaining small databases and catalogues, as well as for network applications.
Since the traffic levels are expected to be low, channel access could be handled by a simple ALOHA protocol. Each station simply transmits whenever it has data to send. If the station does not receive an acknowledgement of the data's having been received within a predetermined time, the data is assumed to have been corrupted or lost, and the station re-transmits it.
In a situation with one large and many small users, a more effective utilization of the radio channel might be achieved by using a slotted aloha protocol. Tannenbaum (1981) provides a substantial discussion of the relative merits of various aloha schemes. The major drawback of any slotted scheme, however is the requirement for all stations to be synchronized, so that all know when time slots begin.
TCP/IP protocols could be used in this model, implementations of which are readily available and inexpensive or free.
The above topology, however, raises a serious issue: the use of relatively high-power transmitters, and the use of already crowded HF or VHF bands will almost certainly require permission and licensing by one or more regulatory agencies.
3.3.2 Multi-hop packet radio networkThe second model, depicted in Figure 3, is a larger, multi-hop network with higher station density at smaller distances. This type of network would be useful where a greater number of network nodes are required and/or the RF frequency necessitates the use of repeaters. The higher density of stations permits a move towards line-of-sight transmission, and thus to higher frequencies. Higher frequencies, in turn, enable improved data transfer rates and improved reliability of transmission.
Use of line-of-sight transmission, however, implies that not all stations may be able to transmit to a central site in a single hop. Assuming a central site is still required for database maintenance (or a gateway to other networks), then some stations will have to send data to the central site (and vice versa) through one or more intermediaries. This use of remote stations as repeaters raises the problem of routing data within the network.
When a packet is transmitted by one station, some or all of the neighbouring stations will receive it. Each receiving station must make a separate decision as to whether or not to rebroadcast the packet. If each station were automatically to rebroadcast, then the packet would bounce back and forth between stations forever. Therefore, stations must have rules to let them decide whether to rebroadcast or not. Tannenbaum (1981) gives a thorough description of this issue, from which much of the following discussion is derived.
In a fixed network, it is feasible for each station to maintain distance information that enables it to know whether it is nearer the packet's destination than the previous transmitter in terms of number of hops. If it is nearer, the station will rebroadcast the packet, changing the identification of last transmitter to that of itself. If it is not nearer, then it will not rebroadcast.
One advantage of this scheme is that the rebroadcasting of the packet can itself serve as an acknowledgment that the packet has successfully travelled another hop. Failure to hear a rebroadcast or acknowledgement within a certain period of time would result in a re-transmission of the original packet. Thus hop-by-hop acknowledgements will not be required, and only end-to-end acknowledgements need be sent.
This routing protocol supports station-to-station traffic as well as station-to-centre traffic as described in the earlier example. A network based on this design will more efficiently support widespread distribution of information resources, and station-to-station messaging than will the previous network.
Equipment to support this network topology is commercially available. One manufacturer offers 2 watt radio modems which have a range of 25-35 km. These modems incorporate physical and data link protocols; network layer and higher protocols will need to be provided by the network implementor. The modems offer a data rate of 9600 bps, sufficient to support image delivery at same rate as commercial group 3 fax machines.
The packet radio portion of this model would be identical to either the 1-hop or multi-hop network outlined above, except that one of the stations would be connected to both the external network and the packet radio network. This common station would act as a gateway between them. Gateways are packet translation devices that interpret addresses at the IP level and assign address that are appropriate for both the local and external network. (Leiner, et al, 1987).
|Latest Revision: April 27, 1995||
Copyright © 1995-2000
International Federation of Library Associations and Institutions