Connecting a local area network to the Internet isn't necessarily hard; it's just a lot of work. What makes it problematic is that it's one of those computer networking problems for which nobody's yet developed a complete, shrink-wrapped solution. Buy this box, load this software, and cyberspace—here we come.
Unfortunately, connecting a LAN to the Internet seems to be one of those tasks that ranks in difficulty with VTAM sysgens for IBM mainframes and determining exactly what all those files called AUTOEXEC.BAT really do. Establishing an Internet connection isn't quite as formidable a task as programming a VCR to record one program while you watch another, but it's close. At least it seems that way to anyone who has tried it. There aren't very many cookbook Internet connection recipes, because everyone seems to start out with a different set of ingre-dients.
One of the problems is that there isn't just one single way to connect your LAN or your host to the Internet. There are several ways to do it, and each depends on the services your users expect to use, how much system configuration you can do, and the type of connection for which you're willing to pay.
So it isn't, unfortunately, a plug-and-play operation. You must do some up-front work to analyze how you're going to configure your connection from your LAN or local host to the Internet, and what to expect once you get there.
First, you must address some basic issues to determine what you're going to get out of your Internet connection. Here are some questions that systems and network administrators must answer before proceeding with an Internet connection:
These questions indicate that there are several alternatives to connecting your network environment to the Internet. Your answers help indicate which type of connection is most appropriate for your situation.
Let's start with an understanding of the basics of connecting to the Internet. The Internet is a group of computer networks connected by data communications circuits, which are phone lines. There are hundreds of separate networks in the group of interconnected networks called the Internet. All you want to do is to connect somewhere to one of them. That one connection is all you need to get users on your LAN into the Internet. It's much like the one connection from your telephone at home that interconnects you to any other telephone in the world. As long as you're connected to one local phone system, you're connected to all of them. An Internet connection works the same way.
So, you need that connection to the Internet. If you're part of a university or a research organization, you may already be part of a network that is one of the Internet's member networks. If you're not, that connection will probably be provided by an Internet access provider. These companies give organizations outside the member network community a port through which they can connect to the Internet. They charge a fee for this service and offer different types of services and capabilities (see Chapter 8, "Finding Access as an Organization").
Things are simpler for military installations that want to connect to the Department of Defense's Defense Information Systems Network (DISN), formerly called the Defense Data Network (DDN). Hosts running TCP/IP, or devices on LANs that run TCP/IP, connect directly to an X.25 packet-switching node (PSN) or to a router that connects to the DISN backbone. The part of the DISN that carries unclassified traffic connects through gateways to the Internet.
It would be convenient if establishing that phone line connection to the Internet were enough, but remember, this isn't an average shrink-wrapped application. One end of the phone line goes to the Internet access provider, but the other end is yours. Your end of the line plugs into a modem or a CSU/DSU. Providing the RS-232 port on your network for the connection to the back end of the modem is another one of those Internet connection problems.
On most networks, you have to configure a router, either in a separate router box or in a host or server configured to act as a router. The router redirects, or routes, traffic from your network out onto the Internet, for delivery to its destination.
The second problem is communicating over that connection to the other hosts on the Internet. An Internet host is a computer system to or from which you transfer files. It can accept logins from remote terminals and run programs. It also can originate, receive, store, or forward electronic mail.
So that all hosts can communicate on the Internet, the use of a standard set of communications protocols has been agreed on so that everyone speaks the same language. That language, or protocol, is the Transmission Control Protocol (TCP) and the Internet Protocol (IP), usually referred to by a single acronym, TCP/IP. The hosts on your network must speak TCP/IP because, by common agreement, all other hosts on the Internet speak TCP/IP.
On the Internet, as on other systems that use this standard set of protocols, TCP/IP implies the use of application-level protocols, such as the File Transfer Protocol (FTP) and a virtual terminal protocol, Telnet. These are two of the most commonly used Internet application protocols, but there are several others.
Your hosts have to run TCP/IP and higher-level applications, or something has to run them for them. If your LAN workstations don't run TCP/IP, a local host can do it for them. Alternatively, your Internet access provider may have a host that can do it for you. Another option is to subscribe to an information service or electronic mail system that gives you Internet access as part of the information or electronic mail access package. This option limits the availability of Internet services, but it allows you to sidestep most of the thorny issues of Internet con-nection.
If you want to connect your LAN-based electronic mail system to the Internet, you have the task of establishing an electronic mail gateway. The gateway translates your local electronic mail address to Internet addresses and then forwards messages to the Internet. A response to your message from a user on the Internet traverses the Internet and is forwarded back to the electronic mail gateway on your LAN. The gateway translates your Internet address back into a local electronic mail system address and delivers the mail. The translation by the gateway is necessary, because your Internet mail address and your LAN electronic mail address are usually different.
The first decision point is to determine how you want to connect your LAN to the Internet. You may be directly or indirectly connected to the Internet. A host or network that is directly connected has full access to all the other networks that make up the Internet. Traditionally, full connectivity to the Internet has been limited to organizations that have some sort of government sponsorship. However, as interest in and use of the Internet has grown, commercial service providers—designed to provide Internet connection for commercial and individual customers—have been permitted to offer direct Internet connections.
An indirectly connected network is a LAN or a separate network that connects to the Internet through another computer system. The intermediary system provides Internet access, but it may allow only limited services, such as only electronic mail.
Directly connected networks need an officially designated Internet Protocol (IP) network address. Hosts on directly connected networks have IP addresses that identify them differently from any other host.
Because you will be connecting your LAN, host, or gateway to a world-wide collection of systems that use the TCP/IP protocols, you need a unique set of IP addresses that identify your systems. You also need an IP address for each host that runs the IP protocol. For example, the IP address 150.48.236.5 uniquely identifies a host running the IP protocol anywhere in the world.
IP addresses are similar to Ethernet network adapter card addresses. Every Ethernet adapter has a unique, 48-bit Ethernet address that identifies it. Similarly, every IP host has a unique IP address that identifies it. The implication of this is that a host or a workstation attached both to a LAN and the Internet has two mutually exclusive addresses. One is for LAN access; the other is for Internet access.
To ensure uniqueness, Ethernet addresses are assigned by the Institute of Electrical and Electronics Engineers (IEEE), which assigns blocks of Ethernet (and Token Ring) numbers to adapter manufacturers. The uniqueness of your IP addresses is established by the InterNIC.
An IP address is a 48-bit binary number expressed as a series of four decimal digits. It is composed of two parts: a network number and a host number. The first part of the number is the network number, and the last part is the number of the host on that network. A host with the IP address 150.48.236.5 is located on network 150.48.0.0, and it is host 0.0.236.5 on that network. In this case, the first two numbers indicate the network number and the last two specify the host on that network. Other types of IP addresses use either the first octet for a network number and the last three octets for a host number, or they use the first three octets as a network number and the last octet for a host number.
Your LAN gateway to the Internet will "advertise" to other hosts on the Internet that it knows how to get to any host on a network for which it acts as an Internet gateway. The other hosts and gateways on the Internet don't know and don't care where a specific host is located. All they need to know is how to route IP datagrams to the host or gateway that "fronts" for a specific network. The gateway then forwards the traffic to the appropriate host.
To connect your LAN and its hosts to the Internet, you must apply to the Internetwork Network Information Center for a network number. Contact the InterNIC for the application and for information about the application process. Its address, phone number, and Internet address are given here:
InterNIC Registration Services
c/o Network Solutions Inc.
505 Huntmar Park Drive
Herndon, VA 22070
1-800-444-4345
hostmaster@internic.net
If you are connecting your own network of hosts to the Internet, you also need a domain name. A domain is an organizational entity that helps identify the hosts in your network by giving them common text names that can be mapped back to IP addresses. The InterNIC also approves requests for domain names after determining that a new domain name is unique.
Establishing a connection to the Internet usually means providing these four primary connectivity elements:
Often, the simplest way to connect to the Internet is to let somebody else do it for you. You can subscribe to an Internet access service that handles all the gory details of Internet access for you. For a monthly use or connection fee, an Internet access provider can give you Internet connectivity across a standard dial-up line. The Internet access provider gives you dial-up access to its host. The Internet access provider's host runs TCP/IP, connects to the Internet, runs the TCP/IP protocols, and routes traffic to and from the Internet. In addition, the access provider usually maintains users' Internet mailboxes and administers mailbox usage.
Reaching the Internet through an access provider carries the extra benefit of little extra expense for hardware, software, or support. Users on a LAN can use standard communications software, such as Crosstalk or Procomm Plus, to dial out of the network through a directly attached modem, or one on a communications server or a modem pool. The most common requirement is that the user's PC emulate a VT-100 terminal (to get full-screen service) but that capability is built into any communications software sold today.
For a relatively low use or connection fee, the Internet access provider lets you have access to all the Internet's electronic mail, file transfer, and host login services. For example, Digital Express (DigEx), of Greenbelt, MD (1-301-220-2020), charges about $20 per month plus a one-time $25 setup charge for dial-up access to its host. The DigEx host gives its users a full Internet connection, a mailbox, and up to 5M of storage space.
Performance Systems International (PSI) of Herndon, VA (1-800-827-7482), offers a similar setup. PSI subscribers use their own PSILink Software to connect to the Internet through local dial-up ports in more than 100 U.S. cities. Users get an Internet mailbox only (PSILink Lite) or a mailbox and FTP (PSILink Basic). Costs range from $9 per month for DOS access to PSI Lite at 9,600 bps, to $39 per month for Windows access to PSI Basic at 14,400 bps, plus a one-time $19 setup fee.
Another Internet access provider, Delphi, of Cambridge, MA (1-800-544-4005), charges $10 to $20 per month for Internet access, depending on how much you intend to use the service and how long you will be connected. Delphi also provides an Internet mailbox and access to the full set of Internet applications, such as FTP and Telnet.
For users who need only electronic mail access, there's an even simpler solution. Commercial online data services and electronic mail systems, such as MCI Mail, Prodigy, and CompuServe, can access the Internet's electronic mail systems through their own electronic mail gateways.
For example, MCI Mail users can dial up the local MCI Mail node on an 800 number (1-800-234-6245), access their MCI Mail mailboxes, and send and receive electronic mail with users on several other electronic mail systems—including the Internet. Normal MCI Mail message charges apply (45 cents for a message of less than 500 characters and $1 for up to 10,000 characters) for any message, even one destined for an Internet address. An MCI Mail user can send an electronic mail message to an Internet user by embedding the recipient's Internet address in the address of the MCI Mail message.
Dial-up access through an Internet access provider can be an expensive solution for an organization with many users who want Internet access, particularly if they are already connected to a LAN. Usage charges can mount quickly, and each user needs a separate account for his or her mailbox.
One solution is to establish a dedicated connection from the LAN to the Internet. The dedicated connection is a communications line to the Internet access provider that gives you access to the Internet. You provide the rest of the connection, including the hosts, router, and TCP/IP applications.
Another solution is to run TCP/IP on PCs and workstations on the LAN and to install an IP router, which is connected to an Internet access provider over a leased line. This is practical for LANs with PCs as workstations and network servers but no host computer, such as a network of PCs and servers on a LAN running Novell NetWare or Banyan Vines.
Let's use a NetWare LAN as an example. The basic configuration is to run a TCP/IP client on each PC on the LAN, giving each PC the capability to direct IP datagrams to the Internet through a router on the LAN. On the Internet, TCP/IP hosts provide the server-side processes to honor client-side FTP, TFTP, and Telnet requests.
Back on the LAN, a NetWare server or a dedicated device acts as a router, forwarding IP datagrams to and from the Internet service provider's access point. In a typical network configuration that has interconnected LANs, you may already have a router (such as a cisco or Wellfleet router) to forward IPX traffic to other NetWare LANs. That same router can route IP traffic to the Internet.
To gain access to the Internet, make arrangements with an Internet access provider to provide a port for your use and acquire the leased line to the access provider's Internet access point. Then configure the router port to route IP traffic (assuming that you have already requested and been assigned an official IP network address by the InterNIC).
Your router acts as a gateway for the devices on the LAN behind it. You have to assign the router port the IP network number of the network for which it will act as a gateway. For example, if you have been assigned a network address of 200.100.50.0, your router acts as the gateway for the PCs on your LAN, which have IP addresses of 200.100.50.1, 200.100.50.2, and so forth.
To run client-side processes, networked PCs can run Novell's LAN Workplace for DOS software or another client-side software program, such as the Clarkson utilities, FTP Software's PC/TCP, or NetManage's Chameleon. This software gives users' PCs access to the standard suite of client-side applications, such as FTP and Telnet. It also links those application-level protocols to the lower-level TCP and IP network protocols.
A workstation using LAN Workplace for DOS uses the TCP/IP protocols instead of NetWare's native SPX/IPX protocols for communications with TCP/IP servers on the Internet. Users' PCs create IP datagrams addressed to the router. The router forwards the IP datagrams to the Internet for delivery to the correct host.
At the same time, the PCs also use the Novell IPX/SPX protocols to communicate with NetWare servers—unless the NetWare servers are configured to run only the TCP/IP protocols. IPX and SPX are efficient protocols for NetWare servers, but if your LAN has both PCs and UNIX systems (such as Sun workstations), you may want to standardize on TCP/IP.
If you are running LAN Workplace for DOS, the key configuration file for Internet access is NET.CFG. This file contains the PC's IP address, as well as other configuration parameters. In most cases, NET.CFG can be run without modification, except for adding the IP address and the address of the default router on the LAN. You may specify only one default router in NET.CFG (it is specified by its IP address). All other routers (if any others exist) are determined dynamically by the Routing Information Protocol (RIP).
PCs running Microsoft's Windows NT Advanced Server are easier to configure for TCP/IP use because TCP/IP is Microsoft's preferred protocol for communications between work-stations and Windows NT servers on different LANs. In the Windows NT world, a PC that can communicate with a Windows NT server runs Windows NT, the Microsoft LAN Manager client for DOS, or Microsoft's upscale Windows for Workgroups (WFW) client. Of course, Windows NT and WFW are fully integrated with Windows. The DOS client loads before Windows and accesses the NT server through the Windows File Manager. This book refers to all three Windows NT clients.
Windows NT clients use Microsoft's NetBIOS extension, NetBEUI, to reach a Windows NT server on its local LAN segment, and TCP/IP to reach Windows NT servers on other LAN segments. Routers running IP connect the LAN segments together. However, because Windows NT clients already run TCP/IP, they are already set up to create IP datagrams that can be routed to the Internet. Network workstations only need application software, such as an FTP or Telnet program, to achieve full Internet access. TCP/IP applications are not part of the Windows NT client software.
Both the Windows NT and the WFW software include Windows utilities to configure a workstation's IP address and to bind them to the adapter through NDIS drivers. In the DOS client, the workstation's IP address and the pointer to the default gateway are parameters in \lanman.dos\protocol.ini. In both the WFW and DOS clients, the NetBIOS names and IP address of Windows NT workstations and Windows NT servers on other networks must be in the LMHOSTS file. If they are not in LMHOSTS, the client has no way of determining the IP address of a host or domain controller on another LAN.
PCs on a LAN can run both TCP/IP and the network operating system's native protocols. Most PCs on a LAN use a network operating system for communications between networked workstations and network servers. Most networked PCs don't use TCP/IP as their native protocol.
For example, Novell's NetWare uses its own proprietary protocols for LAN communications: the Internet Packet Exchange (IPX) and Sequenced Packet Exchange (SPX) protocols. If your networked PCs are to be Internet hosts and have their own IP addresses, they also have to run the TCP/IP protocol stack.
In the days before extended memory managers, running different protocol stacks meant rebooting each time a different protocol was run. Rebooting gave the PC a different network identity because the protocols were bound separately to the PC's network interface card.
However, NIC and NOS vendors, led by 3Com, Microsoft, and Novell, have developed both standard and NOS-specific interfaces between their protocols and NICs. The benefit of these interfaces is that more than one protocol stack can run at the same time, eliminating the need to reboot every time.
For example, 3Com and Microsoft developed the Network Driver Interchange Standard (NDIS), a software specification for a NIC equivalent to an API for network protocols. Instead of writing a driver for a specific NIC, the network software—even TCP/IP—can interface to the standard NDIS NIC driver, which can be bound to several sets of LAN transport and network protocols.
Novell's Open Data Link Exchange Interface (ODI) provides an equivalent capability for the NetWare shell. ODI is the NetWare 3.11 enabler for running both SPX/IPX and TCP/IP at the same time, binding both to the same NIC.
In NetWare 4.x, Novell has added a more sophisticated capability: Virtual Loadable Modules (VLMs). The NetWare 4.x client loads other network protocols on demand. Each protocol is a separate "loadable module" that loads into memory and runs only on demand, instead of remaining in (and using up) memory all the time. A VLM is the client-side version of the NetWare server's NetWare Loadable Modules (NLMs).
Microsoft's Windows NT simplifies PC configuration matters for TCP/IP protocol use. It uses TCP/IP as its native protocol between LANs, relying on NetBEUI for LAN communications on a device's local LAN segment. To round out the top NOS players, Banyan Vines, like NetWare, uses its own proprietary protocols (Vines VIPX and VIP), but offers TCP/IP as a separately configured optional protocol.
Matters are much simpler if a minicomputer or mainframe host (rather than individual PCs) is the TCP/IP host and terminals or PCs on the network connect to the host as terminals. A UNIX host, for example, already runs TCP/IP for most LAN and WAN communications anyway. Users who access the host through terminals or PC terminal sessions have eliminated much of the need for configuring TCP/IP on individual workstations.
Many LANs are composed completely of PCs, workstations, and servers. Traditionally, devices connected to the Internet were host computers, such as DEC VAXes or IBM hosts. Today, many Internet hosts are host computers, and the population has expanded to include DEC Alphas, Sun systems, HP minis, IBM AS/400s, and many others.
The LAN-connected minicomputer or mainframe that acts as both a router and a host is another alternative for Internet access. Unlike a LAN server, the minicomputer can run the UNIX operating system as its native device operating system. In this case, it's not UNIX per se that is important, but what comes with UNIX. Most UNIX operating systems include the TCP/IP protocols, electronic mail process handlers, and FTP and Telnet client and server processes. The system administrator can configure these services for Internet access.
UNIX systems are the most common Internet hosts. They include all the tools to be well-appointed Internet hosts, IP routers, or both. They use TCP/IP as their native protocols. Also, they can create and manage users' mailboxes, act as routers, and run the communications protocols necessary to connect to the Internet.
The system administrator must configure the UNIX system kernel to support TCP/IP processes. Each UNIX system is configured differently, so we'll describe general principles rather than system specifics. The basic UNIX kernel configuration supports networking on a LAN interface (usually Ethernet), an X.25 serial interface, or both. By default, the system is set up as a router to forward IP datagrams to another network interface.
The kernel configuration binds the TCP and IP protocols to the UNIX kernel, but other processes (called daemons in UNIX terminology) must be started to service the other routing and service protocols needed to route traffic and deliver datagrams. For example, the routed and named processes start the Routing Information Protocol (RIP) and the Domain Name Service (DNS) processes, respectively. Another superserver process, inetd, calls processes dynamically as they are needed, such as FTP, Telnet, and rlogin.
Then the IP address and subnetwork mask (if applicable) of each serial or LAN interface are configured. Each network interface has its own IP address, but the IP process also has to know how to interface to a specific network-level process below it, such as X.25 or the IEEE 802.2 Logical Link Control (LLC) layer. The UNIX ifconfig command and its associated parameters are used to configure each network interface.
The ifconfig command also creates a routing table for the network interfaces that have been assigned IP addresses. For a directly connected Internet host, one of the interfaces can be configured to support an Internet data link protocol, such as the Serial Line Interface Protocol (SLIP) or the Point-to-Point Protocol (PPP).
A host with a direct connection to the Internet uses either SLIP or PPP as the protocol on its interface to the Internet. A DISN host, by contrast, uses the DDN standard X.25 protocol. Both SLIP and PPP are wide area network protocols for encapsulating IP datagrams for transmission to the next router or gateway on the Internet. Ethernet, by contrast, is a protocol that can encapsulate IP datagrams for transmission and delivery on a local area network. A router or a gateway forwards the IP datagrams to their destination, encapsulating and de-encapsulating the datagrams in different protocols.
Both SLIP and PPP were created to enable hosts on the Internet to communicate over serial lines. Both SLIP and PPP support asynchronous dial-up and synchronous, private-line transmission. Either can be used for a connection to an Internet access provider, but the one your system uses depends on what your system and the Internet access provider's system support.
SLIP frames datagrams with special characters that specify the beginning and the end of a datagram. It's a simple technique, not unlike that used by Xmodem, to send blocks of characters asynchronously. Developed in the interest of simplicity, SLIP does not do error detection, nor can it support data compression. Its main purpose is to transmit IP datagrams.
PPP is a standard Internet serial line protocol. It was developed to address SLIP's weaknesses. Like other protocols, such as IBM's SDLC, PPP is similar to the High-Level Data Link Control (HDLC) protocol. PPP includes facilities to negotiate connection establishment and termination, and to negotiate connection options. It also can do error detection, so it can guarantee reliable data delivery, regardless of line quality.
If the host and the router are separate devices, the host must know where the gateway is so that it can send IP datagrams to it for transmission to the Internet. So, the host configuration must specify a default gateway to which it directs Internet traffic. On a LAN, the default gateway is the port of the router that is the gateway from the LAN to the Internet. The gateway's IP address is listed as the default gateway. In other words, it's the gateway to which is sent traffic not bound for any IP network address that the host knows about.
If the host connects directly to the Internet, it has to advertise its existence to the rest of the Internet through a standard Internet gateway protocol. Many hosts, even ones directly connected to the Internet, have to advertise only the existence of networks reachable through them.
The most commonly used routing protocol is the Routing Information Protocol (RIP). Systems that connect other networks to the Internet use the Exterior Gateway Protocol, or EGP. This protocol is being replaced by the newer Border Gateway Protocol (BGP). In any case, one of these routing protocols (usually RIP) must be configured so that devices on the network can make routing decisions. In UNIX systems, the gated process runs RIP, BGP, and EGP, and the routing is configured through it.
If terminals on the LAN access the host through a terminal server, configuring the host for Internet access hasn't changed how they use the host. They're still terminals running applications on the host, but now they can FTP through their local host to other hosts on the Internet. PC users can either run client-side FTP and Telnet applications for communications with other Internet hosts or they can connect to their local host by emulating a terminal.
Not all users need or want to transfer files and log on to remote systems across the Internet. For many users, electronic mail is all the Internet access they ever need.
One of the more common requirements for Internet access is to integrate a LAN's electronic mail system and the Internet's extensive electronic mail network. For example, LAN users may send and receive electronic mail using a LAN mail package, such as cc:mail or Microsoft Mail. Despite the problems this may create for system administrators, users want any electronic mail to be accessible through their own local electronic mail system, even users on the Internet. It's done with a LAN electronic mail system's native gateway, or a more general X.400 electronic mail gateway.
An electronic mail gateway translates the address of a LAN electronic mail message into an address that is comprehensible by another electronic mail system, then forwards it for delivery to the foreign mail system. Forwarding it to the other electronic mail system can involve transmitting electronic mail messages across a wide area network (WAN) using a protocol such as X.25. In addition, forwarding electronic mail to the Internet may also require that the gateway use a TCP/IP electronic mail delivery application protocol, such as the Simple Mail Transport Protocol (SMTP).
Many electronic mail system vendors provide optional gateways from their mail systems into other vendors' mail systems. These native gateways are designed to take electronic mail from a specific electronic mail system and convert it into a format comprehensible by another vendor's mail system. For example, cc:mail makes a gateway from cc:mail to Lotus Notes Mail. It helps that both products are from the same company (Lotus), but the electronic mail formats and addressing conventions of cc:mail and Notes Mail are different, and addresses must be translated between the two.
Most electronic mail systems can configure an electronic mail gateway to forward mail to the Internet. The key to making this work is a translation table that translates native electronic mail system addresses into Internet addresses and vice-versa, and a process that executes SMTP. Users embed an Internet address into a native mail address. The gateway identifies the message as one bound to an external mail system, strips off the native mail system address, and forwards it to the external foreign mail system.
For example, cc:mail allows the system administrator (the Postmaster) to set up foreign domains. The cc:mail system knows that anything addressed to a foreign domain belongs elsewhere, such as the Internet. For example, a user may address a cc:mail message to John Smith at {jsmith@falcon.bigco.com}@internet. The Postmaster has configured the cc:mail post office to know that internet is a foreign domain. The cc:mail post office strips off the cc:mail user name (John Smith), and forwards the mail to the Internet for delivery.
Coming the other way, an Internet electronic mail message addressed to the Internet addressee jsmith@falcon.bigco.com may come into the LAN electronic mail system gateway from an Internet access provider. The LAN gateway has to map the user name jsmith to the LAN mail addressee John Smith@BigCo, and deliver it to the appropriate server.
To transmit the electronic mail to the Internet, the gateway or server running a message handler (the Connectivity Manager of Novell's Message Handling Service, or MHS, is an example of one) picks up mail and forwards it to the Internet access provider, or directly to the Internet. If the gateway connects directly to the Internet, the gateway has to act like a standard Internet host. That is, it has to run the TCP/IP protocols, as well as the Simple Mail Transport Protocol (SMTP) to deliver mail to another Internet host.
Another variation of a native mail gateway to the Internet is the X.400 gateway to the Internet. Many commercial electronic mail systems, such as MCI Mail, Sprint Telemail, and CompuServe, transfer electronic mail messages to and from the Internet through X.400 gateways. The X.400 gateway is a convenient way to get electronic mail into and out of a LAN electronic mail system. In addition, it can transfer electronic mail to any other system that has an X.400 gateway.
At the risk of oversimplifying a seriously complicated subject, X.400 is a CCITT standard for addressing and exchanging electronic mail between different electronic mail systems. The address of a mail message is converted to a standard X.400 format and transferred from an electronic mail system gateway to an X.400 Message Transfer Agent (MTA). The MTA transfers the message to another MTA at the destination mail system.
Many WAN carriers operate X.400 MTAs that accept electronic mail messages from a variety of systems with X.400 gateways and then forward X.400 messages to destination systems with their own X.400 gateways. The Internet can be one of the X.400 systems, as can your LAN electronic mail system.
For example, Sprint's Telemail system acts as an X.400 MTA for mail from the Internet destined for other electronic mail systems. If a router on your LAN has an X.25 link to the Sprint Telemail system, mail messages in X.400 format can be directed from the Internet to Sprint, then transferred across the X.25 link to your router. The router transfers the messages to the local X.400 MTA, and the MTA forwards them to the electronic mail system's X.400 gateway for conversion back into the native mail system format.
The TCP and IP protocols provide common communications mechanisms for all users of the Internet. Given the IP address of a destination, the Internet's hosts, gateways, and routers deliver the message.
However, getting the exact IP address can be a problem. Just like dialing a phone number, being close rarely counts. You have to be exact. An IP address is all numbers and can be up to 12 digits long. The correct digits can be hard to remember. Besides, Internet hosts are rarely referred to by number, but by name. We can easily remember that the files we want are on moosehead.bottle.com, or that a correspondent's electronic mail address is cjones@engine.railroad.com.
In each case, the name of an Internet host is specified as part of the mnemonic by which we refer to Internet hosts. For example, Casey Jones' mailbox is on the Internet host engine.railroad.com, and moosehead.bottle.com is the name of a host. However, neither text name is an IP address. If we don't specify the IP address of the host, something has to do it for us.
Fortunately, part of the infrastructure of the Internet is a service that translates text host names into numeric IP addresses. It's the Domain Name Service (DNS), which lives on a host on your local LAN, somewhere on your network, or out on the Internet. DNS provides name-to-IP address translation as a convenience for Internet users. It's the Internet's version of the telephone system's Directory Assistance, or a name service or clearinghouse in a LAN network operating system.
The DNS maintains a set of tables that map host names to IP addresses. For example, an instance of the DNS that knows about the hosts in the railroad.com domain might have an address table that looks like this:
Host |
IP Address |
|
engine.railroad.com |
155.155.10.79 |
|
caboose.railroad.com |
191.207.221.3 |
|
freightcar.railroad.com |
172.157.12.165 |
If a user of an electronic mail application specified that the message was to be delivered to cjones@engine.railroad.com, the electronic mail service has to translate engine.railroad.com to an IP address before it can hand it over to SMTP, TCP, and IP for delivery. If the local host doesn't know the IP address of engine.railroad.com, it can ask the nearest DNS to translate it.
Because it's inefficient to keep asking the DNS to resolve every IP address, most hosts have access to a local table, the hosts table, usually in a file called HOSTS.TXT. The table can be stored on a host on the LAN, or each host or PC can keep its own local hosts table.
So, a request for IP address resolution goes first to the hosts table. If the host name isn't in the hosts table, a request goes out to the host specified as the DNS server for resolution.
DNS servers themselves are arranged in a hierarchy, passing DNS queries to other DNS servers for resolution. At the highest level of the hierarchy are top-level domains. Hosts are arranged into major top-level domains. There may be (and usually are) other levels of domains below each level of the hierarchy.
The host engine.railroad.com is really the host named engine in the railroad domain, which is part of the top-level com domain. There is a DNS server for each level in the hierarchy, and DNS queries are passed to the DNS server at the appropriate level in the hierarchy for resolution.
The last part of the name of most Internet hosts is one of the six top levels of the Internet naming hierarchy for the United States. The top-level domains, and the operators of hosts usually in each, are as follows
Domain Name |
Meaning |
|
.com |
Commercial organizations |
|
.edu |
Universities |
|
.mil |
Department of Defense and other military agencies |
|
.gov |
Government agencies |
|
.net |
Network resources |
|
.org |
Other organizations |
To resolve the IP address of an Internet host from its symbolic name, users' PCs need a pointer to the nearest Domain Name Service (DNS) server or a local copy of the hosts file. With Novell's LAN Workplace for DOS, the RESOLV.CFG file points to the root name server. If the DNS server is not available or can't resolve a name, the inquiry goes to the hosts file in \NET\TCP.
In Microsoft's client software for Windows NT, the name of the host is specified in the initial configuration screens. If you use the standard (and free) DOS client instead of Windows NT as your workstation operating system or Windows for Workgroups, the DNS server address is in \LANMAN.DOS\PROTOCOL.INI. In a standard Windows configuration, the name is in the \WINDOWS\PROTOCOL.INI file.
In each of these implementations, the name of the nearest DNS server is specified as an IP address, rather than a host name. This arrangement avoids the trap of trying to find a DNS host if you only have the name of the DNS server and not its IP address. Therefore, the IP address of the nearest DNS host (as well as a host that has the HOSTS.TXT file) is usually part of the initial system configuration.