40 Point-to-Point Protocols and Links Chap.2 Section 2.8 then treats a number of network layer issues,starting with addressing. End-to-end error recovery,which can be done at either the network or transport layer,is discussed next,along with a general discussion of why error recovery should or should not be done at more than one layer.The section ends with a discussion first of the X.25 network layer standard and next of the Internet Protocol (IP).IP was originally developed to connect the many local area networks in academic and research institutions to the ARPANET and is now a defacto standard for the internet sublayer. A discussion of the transport layer is then presented in Section 2.9.This focuses on the transport protocol,TCP.used on top of IP.The combined use of TCP and IP (usually called TCP/IP)gives us an opportunity to explore some of the practical consequences of particular choices of protocols. The chapter ends with an introduction to broadband integrated service data net- works.In these networks,the physical layer is being implemented as a packet switching network called ATM,an abbreviation for "asynchronous transfer mode."It is interest- ing to compare how ATM performs its various functions with the traditional layers of conventional data networks. 2.2 THE PHYSICAL LAYER:CHANNELS AND MODEMS As discussed in Section 2.1,the virtual channel seen by the data link control (DLC)layer has the function of transporting bits or characters from the DLC module at one end of the link to the module at the other end (see Fig.2.1).In this section we survey how communication channels can be used to accomplish this function.We focus on point-to- point channels(i.e.,channels that connect just two nodes),and postpone consideration of multiaccess channels (i.e.,channels connecting more than two nodes)to Chapter 4.We also focus on communication in one direction,thus ignoring any potential interference between simultaneous transmission in both directions. There are two broad classes of point-to-point channels:digital channels and analog channels.From a black box point of view,a digital channel is simply a bit pipe,with a bit stream as input and output.An analog channel,on the other hand,accepts a waveform (i.e.,an arbitrary function of time)as input and produces a waveform as output.We discuss analog channels first since digital channels are usually implemented on top of an underlying analog structure. A module is required at the input of an analog channel to map the digital data from the DLC module into the waveform sent over the channel.Similarly,a module is required at the receiver to map the received waveform back into digital data.These modules will be referred to as modems (digital data modulator and demodulator).The term modem is used broadly here,not necessarily implying any modulation but simply referring to the required mapping operations. Let s(t)denote the analog channel input as a function of time:s(t)could represent a voltage or current waveform.Similarly,let r(t)represent the voltage or current waveform at the output of the analog channel.The output r(t)is a distorted,delayed,and attenuated version of s(t),and our objective is to gain some intuitive appreciation of how to map40 Point-to-Point Protocols and Links Chap. 2 Section 2.8 then treats a number of network layer issues, starting with addressing. End-to-end error recovery, which can be done at either the network or transport layer, is discussed next, along with a general discussion of why error recovery should or should not be done at more than one layer. The section ends with a discussion first of the X.25 network layer standard and next of the Internet Protocol (IP). IP was original1y developed to connect the many local area networks in academic and research institutions to the ARPANET and is now a defacto standard for the internet sublayer. A discussion of the transport layer is then presented in Section 2.9. This focuses on the transport protocol, TCP, used on top of IP. The combined use of TCP and IP (usual1y cal1ed TCP/IP) gives us an opportunity to explore some of the practical consequences of particular choices of protocols. The chapter ends with an introduction to broadband integrated service data net￾works. In these networks, the physical layer is being implemented as a packet switching network called ATM, an abbreviation for "asynchronous transfer mode." It is interest￾ing to compare how ATM performs its various functions with the traditional layers of conventional data networks. 2.2 THE PHYSICAL LAYER: CHANNELS AND MODEMS As discussed in Section 2.1, the virtual channel seen by the data link control (DLC) layer has the function of transporting bits or characters from the DLC module at one end of the link to the module at the other end (see Fig. 2.1). In this section we survey how communication channels can be used to accomplish this function. We focus on point-to￾point channels (i.e., channels that connect just two nodes), and postpone consideration of multiaccess channels (i.e., channels connecting more than two nodes) to Chapter 4. We also focus on communication in one direction, thus ignoring any potential interference between simultaneous transmission in both directions. There are two broad classes of point-to-point channels: digital channels and analog channels. From a black box point of view, a digital channel is simply a bit pipe, with a bit stream as input and output. An analog channel, on the other hand, accepts a waveform (i.e., an arbitrary function of time) as input and produces a waveform as output. We discuss analog channels first since digital channels are usually implemented on top of an underlying analog structure. A module is required at the input of an analog channel to map the digital data from the DLC module into the waveform sent over the channel. Similarly, a module is required at the receiver to map the received waveform back into digital data. These modules wil1 be referred to as modems (digital data modulator and demodulator). The term modem is used broadly here, not necessarily implying any modulation but simply referring to the required mapping operations. Let set) denote the analog channel input as a function of time; set) could represent a .voltage or current waveform. Similarly, let ret) represent the voltage or current waveform at the output of the analog channel. The output ret) is a distorted, delayed, and attenuated version of set), and our objective is to gain some intuitive appreciation of how to map