Introduction to Internet Networking John M. Wobus Communications & Development Computing & Network Services Syracuse University February 22, 1989 Document Number: IINTRO-1 (c) Syracuse University Computing & Network Services 1989. Copy- ing, in whole or in part, is permitted only for educational pur- poses and copies must include this copyright notice. Copying or republishing for commercial advantage is prohibited. For permis- sion to republish or distribute, write to: Director of Computing & Network Services, Syracuse University, Skytop Office Building, Syracuse NY 13244. ABSTRACT This is a brief, and by no means complete introduction to TCP/IP networks. It includes an introduction for users, some comparison with other types of networks, an introduction of how it works (useful to programmers and network administrators) and recent issues for network administrators. Also included is a glossary. Abstract ii PREFACE An "internet" is a network based on a particular data- communications protocol called IP (for "Internet Protocol"). We use the clumsy phrase "internet networking" to mean the develop- ment and use of such a network (we dislike the term "internet- working" which is sometimes used). We use these terms to distin- guish such networks from those that use other data-communications protocols such as DECnet or SNA. The term "TCP/IP" is also used in various ways to designate internet networking. IP is one of many data-communications protocols for exchanging data between computers. A data-communications protocol repre- sents the rules agreed upon by two or more computers for passing data among themselves. Other data-communications protocols include DECnet, SNA, NJI/NJE, and OSI. IP was invented for use by the Military (for ARPANET), and is also used by research and educational institu- tions, the US Government, and commercial users, especially in the field of computer aided design and graphics. Hardware and soft- ware add-ons to support IP are available for virtually every com- puter from the IBM PC or Macintosh up to the largest supercompu- ters. Virtually every low-level networking technology can be used by IP to link the computers together (serial lines, Ether- net, Token Ring, X.25 packet-switching networks, etc.). Introduction to Internet Networking iii CONTENTS Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . ii Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Users' Introduction . . . . . . . . . . . . . . . . . . . . . . 1 Alternatives to Internet Networking . . . . . . . . . . . . . . 3 Basics of Internet Networking . . . . . . . . . . . . . . . . . 5 Some Necessary Internet Jargon . . . . . . . . . . . . . . . 5 Uniform Addressing . . . . . . . . . . . . . . . . . . . . . 5 Names Versus Routing . . . . . . . . . . . . . . . . . . . . 6 Protocols and Layers . . . . . . . . . . . . . . . . . . . . 7 Internet Protocol Definitions . . . . . . . . . . . . . . . 8 Commentary . . . . . . . . . . . . . . . . . . . . . . . . . 9 Appendix A: Recent Issues . . . . . . . . . . . . . . . . . 10 Qualified Names . . . . . . . . . . . . . . . . . . . . . 10 Domain Name Service . . . . . . . . . . . . . . . . . . . 10 Distributing Routing Information . . . . . . . . . . . . . 11 Routing Example: a University on NYSERNet . . . . . . . 12 From the University . . . . . . . . . . . . . . . . 12 To the University . . . . . . . . . . . . . . . . . 12 Subnetting . . . . . . . . . . . . . . . . . . . . . . . . 13 Broadcast Addresses . . . . . . . . . . . . . . . . . . . 13 Gateway and Host Requirement RFCs . . . . . . . . . . . . 14 Appendix B: Glossary . . . . . . . . . . . . . . . . . . . . 15 Contents iv USERS' INTRODUCTION When you use a computer that is tied to other computers via an internet, you will usually find three basic "network services": Virtual Terminal Service This service gives you the ability to switch your ter- minal from one computer to another. The "internet" term for this service is "TELNET" (different from "Tel- enet", the name of a particular commercial network that has nothing whatsoever to do with internet networking). Typically, you type a command called "TELNET" along with the name of the computer you want to use. It will respond with its "login" prompt, just as if you were directly connected. You can proceed to use that com- puter if you are enrolled as one of its users. File Transfer Service This service gives you the ability to move files between one computer and another. The "internet" term for this service is "FTP". Typically, you type a com- mand called "FTP" along with the name of the computer with which you want to exchange data. You will be asked for a user name and password. Then you have FTP commands that can copy files or display directories of files. Electronic Mail Service This service gives you the ability to send mail elec- tronically to users of other computers on an internet. The "internet" term for this service is "SMTP" though there is typically another name for the command that invokes the service (e.g. "MAIL"). You address an electronic message with the name of a computer and a user identifier, type in the message (or get the text of a message from a file), and send it. The internet software will transmit the mail immediately if it can, or hold it and send it later if the network is not presently working (e.g. due to a faulty telephone line). Personal computers such as IBM PCs or Macintoshes sometimes do not offer all three of these services. TELNET is the most uni- versal, FTP is sometimes replaced by a similar service called TFTP or is only partially offered, and electronic mail is often not offered at all. Efficient electronic mail requires a comput- er which is running nearly continuously, ready to receive mail, so it is typically left for larger computers to handle. There are other, less-universally available services. One important and interesting service is NFS (Network File Service) which gives you the ability to use a disk of another computer on the internet as if it were on your computer. This service requires special "NFS" software for both computers. Introduction to Internet Networking 1 An internet can be small (e.g. two computers) or large (thou- sands of computers). Many institutions participate in a world- wide cooperative internet run on behalf of numerous institutions, called "the Internet". The Internet is used for non-classified military applications, educational applications, and research applications. It originated as a military-sponsored research project in data-communications (ARPANET) and has grown into a high-performance data-communications network used by many educa- tional institutions for collaboration on all kinds of research. Users' Introduction 2 ALTERNATIVES TO INTERNET NETWORKING Internet networking is a "universal" style of networking in that it is supported on virtually every type of computer imagina- ble. Rivals of internet networking are: OSI The "standard" networking protocol for forming networks between unlike computers. OSI stands for "Open Systems Interconnect". Its main disadvantage is that you can't get the hardware and software to make much of a network out of it yet: few computers yet have the necessary software and you cannot get "routers" (dedicated com- puters for interconnecting various small networks with OSI). On the other hand, everything necessary to build large "internet" networks with your existing computers is tested, available, and running at many locations. Huge internets (such as the Internet) demonstrate that it is practical. OSI is not yet near this stage. This fact is often glossed over by people reporting on net- working issues and sometimes even by people preparing procurement specifications. X.25 A protocol that supports large systems of terminals and computers so that any of the terminals can use any of the computers. It is called a "packet switching" pro- tocol because of the technical details of how it moves its data and the term "packet switching" is often used synonymously with X.25 (though in reality, all the pro- tocols listed here switch packets). An X.25 network also offers the ability to move data from a program in one computer to a program in another. All the other internet services are possible in X.25 networks though they are not necessarily available. X.25 is used for commercial timesharing services and networks in the USA (such as Telenet and Tymnet) but is used much more extensively in Europe where there are even publicly-run X.25 networks administered much as public telephone systems--anyone can connect and anyone can get a stan- dard "address". OSI incorporates the X.25 protocol and fills out its services with a fuller standard set of services. SNA IBM's proprietary protocol. Its usual services are roughly like those of X.25. It ranks as the most interesting proprietary network because virtually every computer vendor wants their computers to work well with IBM computers. NJI/NJE Another proprietary protocol of IBM; one they developed and promoted before SNA. Its main service is to trans- mit files between users on different computers in a manner similar to the way internets transfer mail. Mail can be easily sent on this type of network by sim- Introduction to Internet Networking 3 ply by sending it as a file. NJI/NJE networking includes no "terminal" service. It is interesting because many computer vendors offer it for their own computers, it is much less expensive to run than SNA, and it works well. DECnet DEC's proprietary network. It offers roughly the same services as TCP/IP including NFS. Few vendors other than DEC support it. XNS Xerox's proprietary network for local area networks. IP has been adopted as the "native" network protocol for the University of California at Berkeley's version of Unix (much as DECnet is used for DEC's VAX/VMS or SNA for IBM's MVS). A com- puter running Berkeley Unix is a useful tool for managing an internet. It has the fullest set of services and network- management aids and its implementation is the basis for many oth- er implementations. Many desk-top graphics-oriented computers such as Sun workstations have adopted Berkeley Unix as the basis for their software and are designed to use internet networking to exchange data. It is not that easy to set up and manage a network that spans computers of different vendors, using IP or anything else. Run- ning a DECnet network that ties only VAX/VMS systems is much eas- ier than running an internet, but the internet is ready to handle many more different kinds of computers. In fact, some other ven- dors offer DECnet "packages" for their own computers, but as you add more different computers and networking software from differ- ent vendors, the "management complexity" rises in proportion. Alternatives to Internet Networking 4 BASICS OF INTERNET NETWORKING SOME NECESSARY INTERNET JARGON host A computer on an internet with human users. gateway An electronic (computerized) device which connects two or more networks together to form an internet. node A host or a gateway. host/gateway A computer which serves as a host and as a gateway. UNIFORM ADDRESSING The word "internet" was coined to signify a particular kind of network formed by connecting smaller individual networks. IP ("Internet Protocol") is a particular, very explicitly set of conventions for building such a network. The individual networks can be small Ethernets consisting of a few PCs or can be nation- wide packet-switched networks. IP imposes its own "addressing scheme" on the network giving each computer on any of the inter- connected networks its own unique "internet address". Thus each computer must have one address for its individual network and a possibly unrelated internet address. However, internet addresses give you a way to refer to thousands of computers, world-wide on many different kinds of individual networks. Internet addresses consist of two parts: a "network number" and a "host number". The network number designates a individual network. The host number is unique within the individual network and designates a particular computer on it. Internet addresses are always 32-bit numbers and are normally written for human consumption as a list of four numbers in the range 0 to 255 connected with dots, for example: "128.230.1.19". Some of these numbers make up the network number and some make up the host number. Which part is the network number and which is the host number? The part designating the network is to the left, the part desig- nating the host is to the right. Where the division occurs is complicated: it is not always in the same place, but it is always at one of the periods. "Which period?" can be discovered by inspecting the first number: if it's less than 128, then the first period is the dividing point (e.g. 26.0.0.5 divides into network number 26 and host number 0.0.5); if it's between 128 and 191 inclusive, then it's the second period (e.g. 128.230.1.19 divides into network number 128.230 and host number 1.19); and if it is between 192 and 223 inclusive, then it's the third period Introduction to Internet Networking 5 (e.g. 192.26.140.1 divides into network number 192.26.140 and host number 1). Addresses starting with numbers above 223 are reserved for future uses. Why are the addresses so complicated? The scheme manages to handle a few very large networks (millions of computers each) thousands of medium-size networks (thousands of computers each) and millions of small networks (less than 256 computers each). Even so, addresses are compact, being 32 bits (4 bytes) long. In fact, when IP was first conceived, no one dreamed there would be thousands of networks so they really planned for a few large net- works, later adapting it as described above. Where do the numbers come from? An internet must have an administrator who keeps track of network numbers, passing them out to networks as needed. For example, the Internet's network numbers are administered by ARPANET's Network Information Center. Each individual network must also have an administrator who does the same that network's hosts. NAMES VERSUS ROUTING You don't need to memorize internet addresses. Each node has a unique name formed out of letters as well as numbers. You use these name to refer to other computers. However, any network operation can be done without the names; No network operation is carried out by the software until it first looks up all the names to find corresponding addresses. In particular, routing doesn't depend upon names at all. This relationship between the name of an internet computer and an internet address is like the relationship between people's names and their telephone numbers: * I can reach a telephone knowing only a number. You can reach a computer on an internet knowing only its internet address. * If I want to reach a "name" I look up the name in a directory and use the associated number to reach the proper telephone. When you give the name of a computer to an internet service, the first thing the program does is look up the name in a directory to get its internet address. The rest of data routing, etc. is based upon this address. * I can have my own small directory of names of people with their phone numbers, or I can have a published telephone directory with everyone who wants their numbers listed. When you add internet networking to your computer, you can set up a small directory (called the host table) or you can get a huge one from the administrators of the Internet. The direc- tory only lists names and addresses. Routing is something else. All this is hidden from the user, but the administra- tor must understand. Basics of Internet Networking 6 * I can call directory assistance and get the number associated with a name. Internet networking has a similar service, called "domain name service" which computer programs can use in place of a directory (see Appendix for more). Even so, you (the user) only specify a name, you need not worry how the program is finding out the number. PROTOCOLS AND LAYERS Networking "architectures" like internet, OSI, or SNA are organized with protocol layers. The "bottom" layer must include wires and electricity, the physical means to for transmitting data. "Above" this layer is another (with its own set of rules) which makes use of the bottom layer to move its data. Layering aids comprehension of the protocols because the task of moving data reliably from one computer to another throughout a large network is much more complex than one might imagine. A full set of protocols can be divided into many layers: OSI uses 7. Dif- ferent layers have different responsibilities. I'll give exam- ples of such responsibilities below. The "layered organization" used in data communications are really much like the organization of a computer. The bottom lay- er of a computer system is the hardware. Above that is typically a layer called "firmware". On top of that is an operating system which is often composed of two layers: one smaller program to manage processes and some larger programs to provide the the long list of services available. Above that is typically be a run- time package for some language, then an application program. The application program has some sort of input language, another lay- er. A programmer is typically concerned with only one or two of these layers at any one time. Connecting computers together, the main task of data communications, requires that each computer must handle each protocol layer according to exactly the same conventions. Thus, explicit discussions of all the layers. The internet model has four layers. Borrowing the OSI names that correspond most closely: Data-Link This layer consists of "everything below IP". IP is designed to use many different data-link layers, for example, Ethernets or Token Rings. IP can also use services provided by other more advanced networks or protocols to serve as its data-link layer. For exam- ple, an X.25 network can serve as a data-link layer for IP. Network This layer is handled by IP. It takes care of unique addressing throughout the network and the routing of data. Data is carried in units called "datagrams". Each datagram consists of anywhere from one to thou- sands of bytes of data along with the internet address of its destination "control" information. IP does not Introduction to Internet Networking 7 guarantee that datagrams will reach their destination, that they will reach the destination in order, or that they won't be duplicated and delivered twice. This characteristic "unreliability" simplifies the implemen- tation of IP but leaves more to be done in the "higher" layers. Transport This is the layer above IP. Examples are TCP (for Transmission Control Protocol) and UDP (for User Data- gram Protocol). TCP is used for all the standard ser- vices listed above but UDP is used for some more exotic services. TCP takes the potentially bad service that IP offers and "cleans it up", offering the next higher layer a "virtual circuit" so that one computer program can send a long string of bytes to another, assured that they will arrive intact, in order. If IP fails it too much, then TCP simply waits. UDP does none of this, it merely passes on the type of unreliable ser- vice that IP does, perhaps with the advantage of faster delivery. Application This layer is for protocols specific to network servi- ces for the end user. Examples are TELNET, FTP, and SMTP. INTERNET PROTOCOL DEFINITIONS This set of protocols is known as the "internet protocol suite" or "TCP/IP protocol suite". They are defined in two sets of documents. * "RFC"s or "Requests for Comments" are a series of papers which originally served as drafts of the military standards. They are still used for defining IP for the "non-military" world and are still labeled "RFC"s because IP is not an offi- cial standard endorsed by the National Bureau of Standards. Some RFCs really are requests for comments, some are draft definitions of protocols in the internet protocol suite, some are "unofficial standards" for internet networking, and some are guides to implementation and administration. The "unof- ficial standards" are clearly marked as such in an initial paragraph and are enumerated periodically by an RFC titled "Official Internet Protocols". * Military Specifications. Those internet protocols which have been adopted for military use have standard military specifi- cations. Basics of Internet Networking 8 COMMENTARY "Internet" networking seems a bit "experimental". In contrast to other types of networks, the builders of the Internet have spent less time defining the protocols and more time trying them out. The result of this is that internets are doing a lot more at present than other types of networks. Internet protocols have been under constant development to handle new functions and to handle the tremendous growth of the Internet. Many new ideas are tested right on the Internet. The Internet has had its share of problems. These are ascribed to the following reasons: * The Internet is built of many different networks which are administered cooperatively. Effective management and troub- leshooting across administrative boundries can be difficult. * The Internet is formed out of hardware and software designed and build by many different people. The everyday activity of the Internet constantly tests how well all they implement the protocols as well as how well the protocols were defined. * The Internet has grown to a tremendous size, with thousands of computers. The managers of the Internet believe the builders of OSI networks will go through many of the same growing pains. They consider this to be evidence that other networking technologies to serve the same purpose as IP are years behind and could profit by the experience of running the Internet. Introduction to Internet Networking 9 Appendix A RECENT ISSUES QUALIFIED NAMES Most computers are known by a short name like "SUNRISE". This can't be an official Internet name because it is too common. To get around this, official Internet names combine the short name with a set of "qualifiers", strings tacked on to the end of the name with periods in between. For example, the computer known as "SUNRISE" might have an official Internet name of "SUNRISE.ACS.SYR.EDU". These qualifiers categorize the computer as to who administers it and what it is used for--not where it is or how it is reached. Where and how is taken care of by internet addresses and IP- address-based routing (we discuss routing below). If Syracuse University were to create an office in California, it would have no problem giving the office's computer a unique Internet name, since only Syracuse University can use its own qualifiers. Also, a single place like Syracuse University can have two computers on entirely different national networks (e.g. BITNET versus CSNET) but each have names that suggest Syracuse University. The fact that information reaches the two computers by different routes is transparent to the sender. Qualified names were invented to make it easier to cope with the ever-increasing size of the Internet. When they were intro- duced, the administrators of the Internet decided to immediately turn every preexisting name into a qualified name. The name of each node on ARPANET was given the qualifier ".ARPA". Other net- works were treated similarly. The plan assumed that sites would switch to the standard set of qualifiers as soon as possible. However, many people got used to the ".ARPA-type" qualifiers, which have a logic of their own, and their introduction probably slowed the move toward standard qualifiers. DOMAIN NAME SERVICE Domain Name Service is the analogue of telephone directory assistance described above. It is offered by a series of comput- ers (called "name servers") located throughout the Internet which manage a distributed database of computer names. It is designed so that a university (or other institution) can run all the soft- ware on its own computers to manage its own names. Other sites around the world can map this university's names to addresses through automatic queries via the network. Such a query is ini- tially directed to a single name server that passes it along until it arrives at the name server with the internet address required. At each stage, the query is routed by looking at another qualifier. Recent Issues 10 For example, you type TELNET SUNRISE.ACS.SYR.EDU. The TELNET program must find out the corresponding internet address before it can send any data. So, first it sends a query to a name ser- ver about SUNRISE.ACS.SYR.EDU. This name server looks in its own database for the internet address of the name server which han- dles the qualifier ".EDU". It sends the original query on to this name server which in turn looks up ".SYR" in its table. The same thing happens for ".ACS", whose name server has SUNRISE's actual internet address which it returns to your computer, and TELNET proceeds. Remember, the result of such a qeury is an internet address. The program sending the mail (or whatever) hasn't sent any of your data until it has this address. Then it doesn't need the name anymore. The domain name service is a new feature of internet network- ing and is not universally used. Roughly half the world still uses a "telephone book" approach, periodically fetching a name and address table from the administrative site that maintains such a table for the entire Internet. To encourage use of the name domain services, some sites even "unlist" their internet addresses, making them available only through the domain name service. DISTRIBUTING ROUTING INFORMATION In theory, individual computers need not worry too much about how data is routed. Each individual network has at least one internet gateway and each computer on the network has the address of the gateways. When it has data to send, it sends it to one of the gateways which knows where to send it next. Routing between gateways is straightforward: each gateway maintains a list of all network numbers and a parallel list of the gateways to which it should direct data bound for these net- works. This is called the gateway's "routing table". Building and maintaining routing tables is the the hard part. Internet networking doesn't have a comprehensive solution. Here are some of the piecemeal solutions: * All ARPANET gateways have rather complete routing informa- tion. The administrators of ARPANET collect this information and put it in their gateways through their own methods. Oth- er Internet gateways can simply send all data to ARPANET. This simplifies routing tables in non-ARPANET gateways, but turns ARPANET into a bottleneck. So non-ARPANET gateways typically have a table of any routes that they can conven- iently learn, and send to ARPANET only when necessary. To do this, most routing table implementations support the concept of a "default route", a route used whenever a network number shows up which is not otherwise listed in the routing table. Introduction to Internet Networking 11 * Various protocols have been developed for passing routing information between internet gateways. These include RIP, EGP, HELLO, and OSPFIGP. Briefly, each gateway tells all other gateways which networks it is attached to. As this information propagates throughout an internet, gateways can build routing tables. * Any gateway typically has a method for its administrator to add a route to its routing table by hand. This is called a "static route". Routing Example: a University on NYSERNet From the University Gateways within the University can simply send all data to the university's NYSERNet gateway or to the next gateway closer to the NYSERNet gateway. Each NYSERNet gateway has a default route aimed at an ARPANET gateway, it has routes to all the other LANs on NYSERNet from RIP information originating at other NYSERNet gateways, it has routes to NSFnet sites from RIP information that was translated from EGP information by the gateway between NYSER- Net and NSFnet, and it has routes to other NSFnet regional net- works and associated LANs to the extent that these regionals give over their routing information to NSFnet. To the University All nodes on the entire Internet must find out how to reach computers at the university. The administrators of ARPANET have configured the ARPANET gateways to send any data ARPANET receives addressed to university sites in the proper direction. Other Internet gateways use the information passed along via routing protocols. The university uses RIP within its own network to build its own gateways' routing tables. The NYSERNet gateway for the university knows the university network's network number and uses RIP to tell all the other NYSERNet gateways to pass data addressed for the university's network to itself. The gateway between NYSERNet and the NSFnet backbone listens to this RIP and similarly informs all the other gateways of the NSFnet backbone. Other NSFnet regional networks learn to route data to the univer- sity's network towards the NSFnet backbone because the other NSFnet gateways send out the University's network number in what- ever routing information protocol these regional networks are using. Recent Issues 12 SUBNETTING A new scheme divides internet addresses into three parts, the network number, the "subnet" number, and the host number. The purpose of this new division is to allow the internet to treat a collected of small networks as a single network for the purposes of routing. To make this work, any gateway on any of these small networks ("subnets") must know the "true situation" and know how to route between the subnets. Subnet numbers are "squeezed into" the internet addresses by breaking up the normal "host number" into two parts, a "subnet number" and a "host number". All nodes in the subnet must agree upon how much of the address is the subnet number and how much is the host number. Subnetting has become very important to running the Internet as it allows the routing tables to be kept to a manageable size. However, subnetting is a new, special feature of internet net- working and isn't available with every internet software package. BROADCAST ADDRESSES Internet broadcast addresses were devised when LANs were first added to the Internet. LANs usually offer an efficient method of broadcasting data to all computers on the LAN, which can be use- ful in providing some network services efficiently. One of the most important uses is the exchange of routing information between gateways. In order for the network service to function correctly, all nodes on a particular network which contribute to or use the ser- vice must agree on the broadcast address. They have become an issue because people started making them up before an official convention was drawn up. Even though a convention has been devised, a lot of old networking software has not been updated. The most popular convention before the "official" convention remains the only one supported by numerous Unix variants. The "official" convention is an address consisting of 32 bits of ones (255.255.255.255). The convention still followed by some Unix variants is to take the network number and use a host number of all zeros (for network 192.26.137, the broadcast address would be 192.26.137.0). Most new software can be configured to use either of these conventions. The actual choice of convention for each network depends upon whether there are still computers on the network that require the old convention. Introduction to Internet Networking 13 GATEWAY AND HOST REQUIREMENT RFCS There has been a recent effort to clarify exactly how an internet gateway and and internet host should act, to assure that internets function as expected. This has resulted in the comple- tion of an RFC titled "Requirements for Internet Gateways" and the current development of a similar RFC for internet hosts. Ideally, these would codify the best of current practice, but the authors found that they were forced to break some new ground in areas where current practices were causing problems. Unfortu- nately, this means we will have to wait for new versions of net- working software before we gain the benefits of clear definition. Recent Issues 14 Appendix B GLOSSARY address There are two separate uses of this term in internet networking: the phrase "electronic mail address" and the phrase "internet address". An electronic mail address is the string of characters that you must give an electronic mail program to direct a message to a particular person. See "internet address" for its def- inition. ARP Address Resolution Protocol. A protocol run on Ether- nets along with IP to help provide all the networking services necessary to run IP. It maps internet addresses to Ethernet addresses. ARPA Advanced Research Project Administration. The old name of what is now DARPA. ARPANET A data communications network used for communications between research institutions. It is intended for research projects funded by DARPA. Bisynchronous A data-link protocol used by NJI/NJE. BITNET A cooperative data-communications network for colleges and universities that uses NJI/NJE protocols. CSNET A data communications network for institutions doing research in computer science. It uses several differ- ent protocols including some of its own. DARPA Defense Advanced Research Project Administration. DARPA Internet See "Internet". datagram A small collection of data sent through a network as a unit. Many networks provide a "datagram service", the ability to transport datagrams of up to some network- specific size limit to any other computer on the net- work. DDN Defense Data Network. The military portion of the Internet. DECnet A type of data-communications network developed by Digital Equipment Corporations and used mainly to tie together DEC VAX computers. Introduction to Internet Networking 15 default route A routing table entry which is used to direct any data addressed to any network numbers not explicitly listed in the routing table. domain name An internet name that includes qualifiers. domain name server A computer that keeps track of which domain names refer to which internet addresses, and offer this information to other computers on the internet. EGP External Gateway Protocol. A protocol used by the ARPANET to distribute its routing information to com- puters and networks directly connected to ARPANET. Ethernet A type of LAN. It provides only data-link services to attached computers; Further layers are required to make much use of it. FTP File Transfer Protocol. A protocol used on internets to move files from one computer to another. "FTP" also designates the name of the resulting network service and designates the command that invokes the service. GATED (pronounced "gate dee") A program that transfers rout- ing information between various internets connected to it using whatever protocol the network is using to transfer routing information, be it RIP, HELLO, or EGP. GATED is replacement for ROUTED. HELLO A routing information interchange protocol formerly used by NSFnet. host number The part of an internet address that designates which node on the individual network is being addressed. See also "internet address". internet A network that that consists of a collection of smaller networks that communicate between each other using IP. Internet A particular world-wide cooperative internet that includes ARPANET, NSFnet, and various regional, campus, and local networks. internet address A number that designates the interface between a net- work node and a component network of an internet. It has two parts: a network number and a host number. internet name A name that designates a node (a computer or a gateway) on an internet. Glossary 16 internet networking Networking computers using IP. IP Internet Protocol. A protocol which can be used to combine networks into larger networks. It is used on the Internet. LAN Local Area Network. A network that takes advantage of the proximity of computers to offer relatively- efficient higher speed transfers. The hardware and software necessary to put a computer on a LAN is dif- ferent from that required to put it on a wide-area net- work. MILNET A network used for production, but unclassified, mili- tary applications. It is part of the Internet. network number The part of an internet address which designates the network to which the addressed node belongs. Recall that an internet is made up of smaller networks. network service A service offered through a network. Some computers may be "providers" if a network service and some may be "users". Electronic mail is an example of a network service. NFS Network File System. A particular network service that lets a program running on one computer to use data stored on a different computer of the same internet as if it were on its own disk. NFS was invented by the SUN computer corporation. NJI/NJE Network Job Interchange/Network Job Entry. An IBM sty- le of networking based upon their old Bisynchronous protocol. They no longer promote it but it remains popular. NSFnet National Science Foundation Network. A high-speed internet that spans the country. Intended for research applications, it is made up of the NSFnet Backbone and the NSFnet regional networks. It is part of the Inter- net. NSFnet Backbone A network spanning 13 sites across the continental United States. It is the central component of NSFnet. NSFNet Regional Network A network connected to the NSFnet Backbone that covers a region of the United States. NYSERNet New York State Educational and Research Network. An internet to serve NY educational and research institu- Introduction to Internet Networking 17 tions. It serves as the NSFnet regional network for New York State. OSI Open Systems Interconnect. A set of protocols designed to be an international standard method for connecting unlike computers and networks. Europe has done most of the work developing OSI and will probably use it as soon as possible. OSI Reference Model An "outline" of OSI which defines its seven layers and their functions. Sometimes used to help describe other networks. OSPFIGP Open shortest-path first internet gateway protocol. An experimental replacement for RIP. It addresses some problems of RIP and is based upon principles that have been well-tested in non-internet protocols. packet switching Receiving packets of data and routing them to their proper destination. This is the activity of "packet switches", computers dedicated to networking that are interconnected with data communications lines in order to make a network. A packet is a collection of data which is small enough to be buffered briefly in a pack- et switch (no more than a few thousand bits). The term "packet switching" is also used more specifically to mean networking based upon X.25. protocol A set of rules that computers follow to exchange data. Such rules are very explicit since they are designed to be handed to different people who will build and pro- gram computers that work together. Protocols also address the graceful handling of hardware and communi- cations lines failures. qualifier Part of internet names. An internet name consists of words separated by periods. The word after each period is called a qualifier. RFC Request for Comments. IP and all the protocols in the TCP/IP protocol suite are defined in documents called "RFC"s. RIP Routing Interchange Protocol. One protocol which may be used on internets simply to pass routing information between gateways. It is used on may LANs and on some of the NSFnet regional networks including NYSERNet. RLOGIN A service on internets very similar to TELNET. RLOGIN was invented for use between Berkeley Unix systems on the same LAN at a time when TELNET programs didn't pro- vide all the services users wanted. Berkeley plans to phase it out. Glossary 18 ROUTED (pronounced "route dee") A program that runs on Berke- ley Unix and uses RIP to exchange routing data with other nodes on an internet. routing table A table used by internet gateways to direct data in the right direction. It consists of a list of network num- bers, each with the internet address of a neighboring gateway (another gateway on one of the networks our gateway is attached to). serial line A class of data-communications line that allows bits to be moved one at a time from one end to the other. Most communications lines fit this generic definition; The phrase is used more specifically to denote telephone lines and similar data communications lines. SMTP Simple Mail Transfer Protocol. A protocol for moving electronic mail between computers on internets. SNA System Network Architecture. IBM's full-featured data communications protocol. In the past, it has been ori- ented towards large IBM computers, but now has been changed to do more for personal computers. static route An entry in a routing table which is exempt from revi- sion due to data passed on by other gateways (through such protocols as RIP). Typically, a routing table entry added to the gateway's routing table by means of an interactive command. subnet A component network of an internet where this internet is serving as a component network in still a larger internet. "Subnetting" is instrumental in helping the Internet serve as many institutions as it does. subnet number A part of the internet address which designates a sub- net. It is sometimes ignored for the purposes internet routing but other times is used. TCP Transmission Control Protocol. A data communications protocol that delivers streams of bytes from one com- puter program to another on a different computer. It uses IP to move the data between the computers. Telenet The name of a commercial packet-switched network. Term "TELNET" is often confused with it. TELNET A protocol for "virtual terminal service". The word is also used to denote the service and the command that invokes the service. It uses TCP to move the data. Introduction to Internet Networking 19 Token Ring A type of LAN. An example is IBM's LAN and the term "token ring" is often used to denote IBM's version. Unix A popular operating system developed by Bell Laborator- ies. It can be run on most large computers and many single-user computers. VAX A popular type of computer sold by Digital Equipment Corporation. A VAX can be small (for 1 user) or large (for 100s of users). virtual circuit A network service which gives you the effect of a data communications line directly between two computers. Your program can send data by submitting it to the net- work service and it arrives at the other computer in the same order you sent it. Along the way, the network is free to scatter and collect the data based upon which lines are available, but it puts it all back in order before delivering it. virtual terminal service A network service which gives you the a simulated ter- minal on a computer. Assuming you have a terminal on some computer (or you have the main keyboard and moni- tor to that computer), a virtual terminal service allows you to use other computers on the network inter- actively. VMS Digital Equipment Corporation's operating system for the VAX. XNS Xerox Network System. A full-service data communica- tions protocol developed by Xerox. It uses Ethernet to move the data between computers. X.25 A data communications protocol developed to describe how data passes into and out of public data communica- tions networks. The public networks Telenet and Tymnet will "speak" X.25 to your computer if you buy such ser- vice. Glossary 20