Networking Systems

Modern businesses commonly employ at least one local area network (LAN), which connects together several computers, also known as workstations, within a small area. A dedicated file server computer is sometimes used to provide a centralised store of information, each computer then being known as a client workstation. Any point in a network where two or more circuits converge, usually corresponding to the site of a workstation, is known as a node.

Traditionally, such networks use Ethernet wiring, usually with at least one Ethernet hub or concentrator. However, older Mac OS machines are designed for Apple’s LocalTalk network, sometimes confused with the related AppleTalk protocol.

The speed of a network is measured in megabits per second (Mbit/s), although the data transfer rate in MB per second (MB/s) is a more useful figure. The latter is approximately one tenth of the rate in Mbit/s. For example, a 10 Mbit/s Ethernet system carries data at 1 MB/s, although older software and disk drives can knock this down to 100 KB/s or less.

Wireless Fidelity (Wi-Fi)

This lets you create a wireless network with anyone in range of your signals. Public networks that use Wi-Fi, such as Surf and Sip, T-Mobile and Wayport, can be found at airports, shops and conference centres, particularly in the USA.

The following forms of Wi-Fi are currently in use:-

IEEE 802.11

This is the oldest standard, employing the frequency-hopping spread spectrum (FHSS) system. This technology is supported by Windows and offers a transfer rate of 2 Mbit/s.

IEEE 802.11b

As used in Apple’s original AirPort hardware and operating on the 2.4 GHz radio band. The theoretical maximum speed is 11 Mbit/s, giving a data rate about 1 MB/s, but in practice is limited to between 4 and 4.5 Mbit/s or 333 KB/s to 400 KB/s.

The transmitted signal is spread over a broad spectrum using a random pseudo-noise (PN) sequence. At the receiver the data is restored by dividing it by the same PN sequence.

IEEE 802.11g

This system, used in Apple’s AirPort Extreme hardware, runs at a higher theoretical speed of 54 Mbit/s, corresponding to around 5 MB/s. In practice, however, it runs at between 20 and 25 Mbit/s, giving a real data rate of about 2 MB/s.

Although designed to replace the 802.11b standard, this system uses the original Wi-Fi radio band and can therefore inter-operate with older equipment. It employs Orthogonal Frequency Division Multiplexing (OFDM) and three non-overlapping channels. As well as increased speed, it suffers less from reflected signals, which are a problem in built-up areas.

Other Standards

Other systems related to Wi-Fi include:-

IEEE 802.11a

This runs at rates of up to 54 Mbit/s, over twice the speed of 802.11b, although it’s less robust than the latter. Unfortunately, products that employ this standard aren’t compatible with 802.11b devices, since this technology uses a new 5 GHz radio band and eight non-overlapping channels.

IEEE 802.11h

This is the European version of 802.11a, again incompatible with the older Wi-Fi standards.

IEEE 802.11n

Currently under development, this offers rates in excess of 100 Mbit/s.

Constructing a Network

A Wi-Fi network is created by placing two or more suitably-equipped computers in range of each other, usually at a distance of less than 150 feet or around 45 to 50 metres. A wireless hub, such as Apple’s Base Station, can be used to extend the range. This device must be placed midway between the computers, allowing it to act as a bridge between the machines.

The range of a wireless hub also 150 feet, although this can sometimes be extended by fitting the hub with an external aerial, giving a range of around 250 feet (75 metres) in all directions, or by using a directional aerial, giving 500 feet (150 metres) in a particular direction: with some hardware you may even reach a distance of 750 feet (230 metres).

Wireless Gateways

A wireless gateway or wireless access point is a device that extends an Ethernet network via Wi-Fi to one or more computers, but isn’t actually a hub. An Ethernet router that contains a wireless port can be used instead of such a gateway.

Another option is a wireless bridge, which joins together separate Ethernet and Wi-Fi networks to create a single system. Similarly, up to four AirPort Extreme Base Stations (see below) can be used to bridge the ‘gaps’ in a wired network. In this instance, one unit, usually the one connected to the Internet, is designated as the master, whilst the others behave as satellites. Apple’s proprietary Wireless Distribution System (WDS) is employed to operate this system.

AirPort and AirPort Extreme

Apple’s implementation of IEEE 802.11b is known as Airport, whilst IEEE 802.11g is referred to as AirPort Extreme. You can only gain access to a Wi-Fi network from a Mac OS machine if it has inbuilt support for AirPort or has an AirPort card fitted in its AirPort slot. However, you can fit a third-party Wi-Fi card into a spare PCI slot or PC Card slot, although older cards only work at 2 Mbit/s, often require encryption to be disabled when using a Base Station and need special driver software. Third-party cards for IEEE 802.11g may also require you to ‘hack’ the standard AirPort software.

The AirPort Base Station

Although AirPort lets suitably-equipped computers communicate directly, you can also employ Apple’s Base Station as a wireless hub. The original AirPort 1.0 software accommodates 10 machines, whilst AirPort 2.0 and later versions accommodate up to 50, although you should only connect between 10 and 20 to maintain sufficient bandwidth.

The original 802.11b Base Station is shown below.

Version 2 of the Base Station has a 10/100Base-T Ethernet port as well as a standard 10Base-T socket for a wideband modem, as required for an ADSL and other broadband connections.

Various types of AirPort Extreme Base Station have been produced, the basic versions having a USB port suitable for USB printer sharing, as well as two 10/100 Base-T Ethernet ports, whilst the bigger models include a 56 kbit/s modem. All of these models can act as a hub, providing optional access to a an Ethernet network and giving several users simultaneous access to a modem, although speed can be compromised if too many people use the latter at the same time.

Each computer is automatically given an IP address for the Internet, using Dynamic Host Configuration Protocol (DHCP), although the Base Station itself appears as a single device on the Internet by means of Network Address Translation (NAT).

Configuring and Using a Base Station

In the Classic Mac OS, most AirPort options are set with the AirPort Setup Assistant application, using settings in the Remote Access and TCP/IP control panels. The Base Station itself is configured using the AirPort Admin Utility.

AirPort Express

Apple’s AirPort Express device is similar to a Base Station but much smaller. As well as Ethernet and USB connections, it provides an audio output via a mini-jack, which can also convey digital audio via TOSlink optical cables, allowing computer-recorded sound to be sent to other equipment via Wi-Fi. It also works with Windows 2000/XP computers and can extend an existing Wi-Fi network in conjunction with Apple Base Stations, but not with third-party Wi-Fi routers.


Ethernet is an industry-standard family of networking systems that run at speeds of between 10 Mbit/s and 10 Gbit/s.

Modern Ethernet Systems

All Ethernet systems operate as a baseband network and run at a specified speed. The speed rating itself is hidden in the code for the Ethernet system, as shown in this table, which includes most of the modern formats.

Max ​Length
10 Mbit/s
100 m
100 Mbit/s
100 m
Fast (FO)
100 Mbit/s
1 Gbit/s
100 m
Gigabit (FO)
1 Gbit/s
10 Gigabit
10 Gbit/s
DT*15 m

UTP = Unshielded Twisted Pair

FO = Fibre-Optic

DT = Dual Twinaxial

* Special Connector

Most modern equipment uses 100Base-T or 1000Base-T, which uses the same connector as 10Base-T. Fortunately, many devices incorporate automatic speed sensing, allowing you to interconnect 10Base-T, 100Base-T or 1000Base-T equipment, although only the more recent devices accommodate the latter standard.

Most forms of Ethernet use unshielded twisted-pair (UTP) ‘telephone’ cable, containing four or eight wires as two or four twisted pairs. An 8-way RJ45 modular plug is used, as illustrated below. Modern computers usually have matching socket, often accommodating both 10Base-T and 100Base-T systems, whilst some machines also provide for 1000Base-T.

Network Hardware

Special hardware is often required to create a complete network. The 10Base-T, 100Base-T and 1000Base-T versions of an Ethernet network are usually in a star form, with a central Ethernet hub, switch or concentrator to which all devices are wired. The functions of a hub or switch are often performed by a router, which uses IP addresses to direct the data to specific destinations. Finally, there’s a bridge, which connects one kind of network to another,


A hub, also known as an unswitched hub, unmanaged hub or dumb hub, simply transfers data as requested. This means that the available bandwidth of your Ethernet system is shared amongst all the devices on the network, reducing the maximum speed of data transfer. Modern hubs often have automatic Ethernet speed sensing.


A switch, also known as a switched hub, managed hub or smart hub, is an ‘intelligent hub’ that creates a faster network. It remembers the address of each destination that receives data, thereby giving every device the full network bandwidth. Switches are also better than hubs at accommodating mixed Ethernet speeds and often incorporate a remote control feature.


A bridge connects one type of network to another, as in an Ethernet to LocalTalk adaptor, which connects an Ethernet network to an Apple LocalTalk network or a LocalTalk printer. The connection between such an adaptor and a hub, switch or router requires a standard Ethernet cable, whilst a connection to a computer may require a crossover cable.


A router is really a ‘cross’ between a hub and a bridge, usually fitted with several Ethernet connections for computers and a single Ethernet socket for wiring to a local area network (LAN). Unlike a bridge, a router employs IP addresses, often in conjunction with Dynamic Host Configuration Protocol (DHCP). This means that a router, unlike many of the other devices described here, can’t convey non-standard protocols, such as AppleTalk, although it accepts AppleTalk over IP.

Ethernet Wiring

When planning a network, you should carefully consider cable distances and allow for future upgrading. For example, the cable length from each device to a hub must always be less than 110 metres. And Category 5 (Cat 5) cable must be employed if you intend to use 100Base-T in the future, thereby avoiding the need to renew wiring in the event of an upgrade.

If you want to use 1000Base-T in the future, you must connect all eight wires in every cable. Many older 10Base-T or 100Base-T systems have only four wires connected, although the unconnected spare wires may still be useable.

Making Ethernet Cables

Although ready-made network cables are available, you can also assemble your own using an Ethernet crimping tool. This is designed for use with the insulation displacement connector (IDC) variety of RJ45 plug. You should proceed as follows:-

  1. Use the crimping tool to square-off the ends of the cable.
  2. Remove half an inch of the outer jacket, avoiding damage to the wires. Untwist the pairs of wires into the following sequence:-
    Blue/White, Blue, Green/White, Green, Brown/White, Brown, Orange/White, Orange
  3. Use the crimping tool to cut the wires straight, push them into the connector and use the tool to attach the cable.
  4. Repeat with the other end of the cable, ensuring that the wires and both plugs have the same orientation.

In fact, the wiring sequence shown above doesn’t seem to be entirely correct as most commercial Category 5 plugs are wired to the following convention:-

1Orange/White *
2Orange *
3Green/White *
6Green *
* Wires for Category 3 cable (10Base-T or 100Base-T only)

Crossover Cables

A crossover cable is sometimes required if you want to connect two devices without using a hub. This requires pairs of wires to be swapped at one of the connectors or a crossover adaptor to be interposed between two standard Ethernet cables. The latter can be obtained from a computer store, but if you want to make one yourself it can be wired as shown below:-

Connector ​1WireColour
Pin ​11Orange/White
Pin ​22Orange
Pin ​33Green/White
Pin ​66Green
Connector ​2WireColour
Pin ​13Green/White
Pin ​26Green
Pin ​31Orange/White
Pin ​62Orange

Fitting Ethernet to an Older Computer

Modern computers usually have a built-in RJ45 connector for Ethernet. However, some older machines have an alternative connector that accepts an external transceiver (see below).

If your machine doesn’t have any provision for Ethernet, you must install an Ethernet card or other special hardware. A matching Ethernet card can be fitted in the following types of expansion slot:-

Whatever Ethernet card or adaptor you use, it must of course match the type of slot in your machine.

If you have an older machine without an expansion slot you may be able to employ an external SCSI to Ethernet adaptor. This plugs into the computer’s SCSI port, but lets you continue using other SCSI devices already connected to the port.

Ethernet Hardware Connections

A computer or Ethernet card may provide a direct Ethernet connection or one that needs an external transceiver. Most Ethernet hardware provides at least one of the following connections:-

You can’t use a computer or Ethernet card that doesn’t have a RJ45, AAUI or AUI connector on a twisted-pair Ethernet network.


You must install an external transceiver if your computer or Ethernet card only has an AAUI or MAU socket . This device, also known as a media access unit (MAU) or access unit (AU), goes between your Ethernet hardware and the network wiring.

The transceiver for older Mac OS equipment is known as an Ethernet media adaptor and has an Apple Attachment Unit Interface (AAUI) connector. The AAUI-15 plug should match the connector on your computer or Ethernet card.

Transceivers designed for other computers have an access unit interface (AUI) connector, usually a 15 way D plug (DB15). This is incompatible with the AAUI connector, requiring a matching Ethernet card, usually of a type designed for a PC. In theory at least, it should be possible to use this kind of card and associated transceiver with a PCI-equipped Mac OS computer.

Different transceivers are used for different forms of Ethernet, allowing you to change the network without replacing the actual Ethernet card. For compatibility with other modern Ethernet devices you should always use a twisted-pair transceiver, sometimes known as a twisted pair AU (TPAU).


FireWire isn’t designed for networking but can be used to create a very fast system. Data is usually conveyed via TCP/IP over FireWire, as supported by Mac OS X 10.3 or higher, although special software such as FireNet can also be used.

A FireWire network can easily be created between two computers by plugging them together. For a larger network you’ll require a FireWire hub, sometimes with one or more FireWire repeaters to cover longer distances. In Mac OS X you can then open the Network preferences pane, click on Configure and choose Network Port Configuration in the Show menu. Now click on New, enter a name, such as FireWire, and select Built-in FireWire from the Port menu.


LocalTalk is an obsolete networking system, as employed in ‘classic’ Mac OS computers. It runs at 230.4 kbit/s, giving an actual transfer rate of around 10 KB/s, which is very slow by modern standards.

Each LocalTalk network or segment, accommodates up to 32 devices, also known as nodes. A network can be expanded to 254 nodes by adding extra segments via one or more bridge devices.

In most instances, a LocalTalk connection is made by plugging a LocalTalk box into the Modem port or Printer port of a ‘classic’ computer. With Classic AppleTalk, as found in older versions of the Classic Mac OS, you must use the Printer port, which is convenient if you have one or more LocalTalk printers. However, the modern form of AppleTalk, also known as Open Transport, lets you employ either port, except on a PowerBook 190 where the Modem port must be used.

LocalTalk Boxes

Each LocalTalk box contains a hybrid transformer. This converts the four-wire RS-422 circuits of a Mac-compatible serial port into a two-wire bidirectional signal. Data in this form can then be carried over the network. A pair of special 3-way mini-DIN connectors, also known as LocalTalk connectors, are provided, enabling the boxes to be connected together in a daisy-chain.

The network is wired using shielded twisted-pair (STP) cable with a characteristic impedance of 75 ohms, the screen joined to the connector’s middle pin. Both LocalTalk sockets on the box have plug-actuated switches, providing automatic termination when a plug is removed, allowing the box to be unplugged from the computer without disrupting the network.

Unfortunately, the LocalTalk plugs do have a habit of falling out. There are two ways around the problem. You can tie the boxes to your office furniture or you can use PhoneNet (see below).


Farallon’s PhoneNet is an alternative form of LocalTalk that employs RJ11 modular connectors (as used for telephones), standard telephone cable and a different electrical system, although it’s still LocalTalk as far as the devices are concerned. Up to 30 Mac OS machines can be connected over a distance of 150 metres, or 450 metres if screened cable is used.

A linear PhoneNet distribution is created by attaching a special PhoneNet box to each device. This also has a plug that engages with the serial port of a device. However, unlike a normal LocalTalk box, this has two RJ11 sockets for the actual network cables. Several such boxes can then be linked together using 2-pair (4-wire) telephone cable, all connected to RJ11 plugs.

In a star distribution, each device is connected via a PhoneNet adaptor and cable, often via a wall plate with an RJ11 socket, to a PhoneNet hub. Devices can be up to 450 metres apart, whilst even longer distances can be covered using repeater boxes.

You can connect a PhoneNet network to LocalTalk wiring by using a cable with a 3-way mini-DIN plug at one end and a RJ11 plug at the other. The mini-DIN plug can be connected to any LocalTalk box or joined to an existing LocalTalk cable by means of a standard LocalTalk coupler.

Using Existing Phone Lines for LocalTalk

In theory, you can use spare wires in existing telephone wiring for LocalTalk, although this could be frowned on by your telephone company. Indeed, in the United Kingdom the connection of any unapproved device is illegal. Furthermore, the British wiring system uses three of the four wires usually found in each cable, making it unsuitable for LocalTalk.

However, in the USA, only two of the four wires in a telephone cable are actually used for phones. Standard phone jacks have a red and green ‘pair’ of wires, usually connected to the phone, as well as a spare yellow and black ‘pair’ that can be utilised for LocalTalk. These can be wired to a additional RJ11 phone jack, using ‘opposite’ pins to those employed for voice circuits. Any such jacks that aren’t connected to a LocalTalk adaptor should be terminated by means of a 120 ohm resistor.

LocalTalk Bridges

An Ethernet to LocalTalk adaptor lets you connect a LocalTalk or PhoneNet network to an Ethernet network, or to an Ethernet-capable printer or computer. A standard Ethernet cable should be used to connect to an Ethernet hub, although a connection to a computer may require a crossover cable.

If your Mac OS computer doesn’t have an Ethernet port but has USB, you can use a USB to LocalTalk adaptor to connect the machine to a LocalTalk or PhoneNet network.

Other Networking Options

There are several other networking systems available, some of which haven’t yet reached their full potential. The possible options include:-

AnyPoint Home Network

This simple type of network, designed for domestic use by Intel, is based on a system developed by the Home Phoneline Networking Alliance (Home PNA). It lets you create a network over your normal telephone lines without upsetting phone calls or the use of a modem or ADSL. It also gives several computers access to the Internet using a common Internet connection.

The original form of PNA runs at 1 Mbit/s, although version 2.0 operates at up to 10 Mbit/s. Unfortunately, version 1.0 is also incompatible with the telephone regulations in the United Kingdom. PNA works by means of special adaptors that plug into your phone sockets, allowing a conventional telephone and computer device to be connected and used at the same time.

Home Phone Networking can be accommodated by fitting PCI card in your computer, which has a standard phone jack for the network wiring. Or you can use an external USB adaptor, connected between a USB port on your computer and a phone jack.


This is a medium-range radio frequency (RF) system that offers a transfer rate of 1.6 Mbit/s. The newer HomeRF 2.0 should provide speeds of up to 10 Mbit/s.

Other Wireless Systems

Several other non-cable networks are available, including HyperLAN/2 and OpenAir. However, it’s unlikely that several competing systems can survive.

Using a Direct Connection

It’s possible to create a small ‘network’ of two computers, or one computer and a printer, by plugging them directly together, using Ethernet, FireWire, USB or, with older machines, LocalTalk.

For an Ethernet connection you may need an Ethernet crossover cable, although some devices work perfectly with a standard cable. A crossover cable can be obtained from most computer suppliers, although you could make your own, as described in the Ethernet section of this document, or you can insert a crossover adaptor between two standard Ethernet cables.

A FireWire network can be created between two computers by simply plugging the machines together with a standard FireWire cable. It’s also possible to use two USB-equipped Macs for peer-to-peer communication by using special software such as Lindy’s USBLink. The author assumes that you could use a specially-wired USB cable with a Type A plug at both ends.

To connect two LocalTalk machines you can use a standard printer cable: LocalTalk boxes and cables aren’t required for such a simple connection.

Ethernet Reference

The full range of common Ethernet systems are as follows:-

Max ​Length
Thick †
10 Mbit/s
0.5" ​CXUHFLinear
500 m
Thin †
10 Mbit/s
0.2" ​CXBNCLinear
300 m
10 Mbit/s
100 m
100 Mbit/s
100 m
Fast (FO)
100 Mbit/s
1 Gbit/s
100 m
Gigabit (FO)
1 Gbit/s
10 Gigabit
10 Gbit/s
DT*15 m

Name is derived from thickness of coaxial cable

CX = Single-core coaxial

UTP = Unshielded Twisted Pair

FO = Fibre-Optic

DT = Dual Twinaxial

* Special Connector

The more usual varieties are discussed earlier in this document. The remainder are less common, although the method of connecting such networks to your computer is often similar. Further details appear below.


This older daisy-chain network is found in large systems. It uses 50 ohm co-axial cable, marked at 2.5 metre intervals to show where a tap can be made for a connection. Each segment of a network must have a length of 500 metres or less, can accommodate up to 100 taps, and must have terminators at each end. Longer distances require a repeater or similar device.

Each tap must be attached to a suitable transceiver, also known as a media access unit (MAU) or access unit (AU). If your Ethernet card doesn’t have an internal transceiver, providing access via a single UHF connector, you’ll need an external transceiver to connect to the network.

The limits on tap positioning are awkward when devices are closely spaced, although such devices can be connected via a single transceiver. For terminals, hosts or modems with slow asynchronous ports you can use a terminal server, also known as a network interface unit (NIU). Or you can use a fan-out box, which has an AUI cables for each device. Since this kind of box also works as a self-contained network without a final transceiver it’s often known as Ethernet-in-a-Box. The final and best option is probably a multi-port transceiver, which can be cascaded to other similar transceivers for extra circuits.

A network can be expanded into further segments by means of repeaters. A multi-port repeater can add several extra segments, using other kinds of Ethernet if required. Long distances can be accommodated by interposing a fibre-optic link between two or more segments.


This alternative daisy-chain system is sometimes known as Cheapernet. Each segment must have a length of 185 metres or less and can have up to 30 connections. A terminator is required at each end of the network, although some devices provide auto-termination to avoid this complication. Longer distances require a repeater, similar to that used for 10Base-5. The network uses 50 ohm co-axial cable and connections to devices can be made at 5 metre intervals.

As with other forms of Ethernet, you may or may not need an external transceiver. If your Ethernet card has a single BNC connector you can connect it to a network via a BNCTjunction adaptor. The integrity of your network is maintained even if a BNC plug is removed from a computer, Ethernet card or transceiver, as long as the adaptor itself isn’t dismantled.


This variation of Fast Ethernet uses fibre-optic cable, which is ideal for connecting local concentrators to a central concentrator, as it isn’t affected by any interference created by differences in the electrical potential between buildings.

If your Ethernet card lacks a direct 100Base-FX connection, you’ll have to plug a special fibre-optic transceiver (FOT) into its AAUI or AUI connector.


This fibre-optic predecessor of 1000Base-T offers a similar speed to some varieties of wide area network (WAN), including ATM and FDDI (see below). If your machine doesn’t have a 1000Base-SX port you’ll need to install a suitable card into a spare slot. However, most users prefer 1000Base-T, since this easily connects into an existing 10Base-T or 100Base-T network.


Defined in the recent IEEE 802.3ak standard, this transfers data at 10 Gbit/s, although only over a distance of 15 metres. It needs dual twinaxial cable and matching connectors, so won’t replace existing systems immediately.

Other Derivatives

Other variants of the IEEE 802.3 Ethernet standard are employed, using various network operating systems, including Novell Netware. Common variations include DECNet, a well-established system based on the 802.3 standard, and Starlan.

Starlan conforms to the 1Base-5 format, operating at 1 Mbit/s over two twisted pairs. It uses RJ45 connectors and a central hub that accommodates up to 12 devices, each with a cable up to 244 metres in length. Up to 10 devices can also be daisy-chained over a distance of 122 metres on one arm of the star. However, such an arm must end in a star-term terminator.

Network Reference

The OSI Protocol Model

The International Standards Organisation (ISO) has devised a seven layer protocol model for Open System Interconnection (OSI). Each layer is built upon those beneath as follows:-

7: ​ApplicationAllows ​recipient ​to ​use ​data ​with ​applications
6: ​PresentationModifies ​files ​to ​suit ​recipient’s ​computer
5: ​SessionCo-ordinates ​actions ​of ​sender ​and ​recipient
4: ​TransportConfirms ​data ​is ​sent ​correctly ​(not ​always ​used)
3: ​NetworkDirects ​data ​to ​the ​correct ​recipient ​(network ​only)
2: ​Data ​LinkBasic ​control ​of ​data ​flow ​and ​error ​detection
1: ​PhysicalDefines ​connection ​system ​and ​speed

Layers 1 to 5 are concerned with internetworking whilst 6 and 7 provide inter-operation, Although the model isn’t fully implemented, various standards for Layers 1 to 4 are established. The remaining layers are more sophisticated and are being developed above the existing layers. Of these, the Presentation Layer is important since it includes file interpretation, decryption and decompression.


The topography of a network describes the route taken for data to pass from one device to another over a network. In many instances this corresponds to the physical layout of the cables used to connect the devices or nodes in the network. The following systems are commonly used:-

Linear Daisy-Chain

In this kind of network, also known as a series network, the devices are connected by a single cable, so one lost connection stops the whole network. All devices share a common circuit, but each device has a unique hardware address.

Linear Bus

Similar to a daisy-chain network but with each device wired via a branch. This means that the loss of a single connection doesn’t necessarily harm the whole network.


In this system individual circuits to each device originate from a central switching hub or concentrator. No electrical circuits are shared, making it more expensive but highly reliable.


A unusual system, similar to a daisy-chain, but with the ends joined to form a circle. Each device extracts the data it needs, regenerates the signal and then passes it on to the next device.

The topography isn’t always obvious. For example, a Token Ring, with each terminal wired to a multi-station access unit (MAU), looks like a star network, even though it’s wired as a ring. In addition, the logical layout needn’t match the physical layout. For example, a Token Bus network is wired in linear form but the data passes between the terminals as in a ring.

Cable Types

Cables usually come in the following forms:-

Unshielded Twisted-pair (UTP)

Originally used for rates of up to 10 Mbit/s, but now extended to 1000 Mbit/s. Since there’s no shielding this kind of cable can cause, or suffer from, electrical interference problems.

Shielded Twisted-pair (STP)

Often used for 16 Mbit/s or higher, without needing regeneration over distances of between 100 metres and 1.5 km.

Co-axial (Coax)

Commonly used for speeds of 50 Mbit/s or higher, without requiring regeneration for 500 metres or more.

Fibre-optic (FO)

Often used for speeds of 100 Mbit/s or higher, without requiring regeneration for 2 km or more. Note that the Bayonet ST connector is more efficient than the 9 mm SMA version.

Transmission Techniques

Most networks operate as a baseband system, providing only one data channel. This means that two devices must be prevented from sending data to the channel at the same time. Different types of network use different approaches to this problem.

Ethernet and LocalTalk use Carrier Sense Multiple Access with Collision Detection (CSMA/CD), in which any device already using the network is given priority while other devices sense the presence of its carrier and have to wait. When the channel becomes vacant it’s possible that two devices will begin transmission at the same time. If this happens they must both stop sending signals, wait a random period of time, and then try again.

Token Bus networks use token passing to effectively provide time division multiplexing (TDM), in which each device has a time slot, but only if it wants to send a message. In a Token Ring network one device acts as a monitor station, issuing tokens that allow other devices to transmit, whilst all the other devices wait to see if they can become the monitor station.

Repeaters and Other Devices

A repeater operates in the Physical Layer, cleaning up the data from one segment of a network and passing it on to another segment. A multi-port repeater is similar, but provides this facility for several segments of a network. Some devices also introduce retiming to compensate for long cables or can reconstruct the preamble in a data packet should it be damaged.

A bridge operates in the OSI Physical and Data Link Layers, although in IEEE 802.x systems it works across the Media Access Control (MAC) and 802.x Physical layers. Such a device provides a store and forward mechanism, selectively passing data, often between networks of a different type.

A router also connects two or more networks of the same or different type, receiving packets of data and sending them onwards, using information within the packet itself to determine the next destination. Internet Protocol (IP), part of the Network Layer, is normally employed, using addresses that are independent of hardware addresses. A router can often assign addresses to each device automatically, in which case it’s known as a dynamic IP address. The most common mechanism used for assigning addresses is Dynamic Host Configuration Protocol (DHCP).

A brouter acts as a bridge until it receives data of a specific type that causes it to use IP in the same way as a router. A gateway operates at any of the higher Layers, perhaps offering some inter-operation between different types of network.


Every Ethernet card or interface has a built-in 48-bit hardware address, also known as a Media Access Control (MAC) address. This identifies each device on a worldwide basis, but means that changing the hardware also changes the address. In some self-contained networks this problem is avoided by giving each station a fixed 16-bit address, whilst Internet-based systems employ a 32-bit IP address to identify each location. Whatever mechanism is used, an address whose lower bits are ‘on’ can be used to broadcast a message to all stations or to multicast it to selected stations.

Older Mac OS computers made by Apple are assigned MAC addresses in two ranges:-



where the digits xx are defined in each device and all the numbers are given in hexadecimal notation. Note however, that modern Mac OS machines employ other ranges, as in the author’s computer, which has an address of

Transmitted data always includes a two-byte length field, although this isn’t required for a fixed-length LLC message. In this situation, it’s often used to indicate which high-level protocol has been employed. In most instances this is TCP/IP, the standard pair of protocols used on the Internet.

Local Area Networks

A local area network (LAN) can be provided over Ethernet, although other types of network can be encountered, such as:-

Fibre Channel (FC)

This is an accepted ANSI standard, employing Fibre Channel Protocol (FCP) to transfer data via a fibre-optical cable and convenient coaxial connectors. In the format known as Fibre Channel-Arbitrated Loop (FC-AL) it accommodates up to 126 devices, usually in the form of high-performance hard disk drives. Each device can be up to 30 metres apart.

The transfer rate ranges from 100 MB/s to a maximum of 200 MB/s, although this may be increased to 400 MB/s in the future. Although very fast, Fibre Channel suffers from the signal delay problems common to all optical systems.

Token Bus

Token Bus, once popular with educational establishments, employs tokens that enable terminals to send data. The physical wiring is in linear or bus form, although data travels in a logical ring. It uses broadband or carrierband techniques, operating at 1 to 10 Mbit/s over co-axial cable.

For this network you must have a TokenTalk card in your computer. Older versions of the Classic Mac OS include suitable software, so you can select TokenTalk (Apple’s implementation of Token Bus) in the AppleTalk or Network control panels.

Arcnet is a star form of Token Bus, once produced by Tandy under the name Vianet. It runs at 2.5 Mbit/s with up to 255 nodes, each of which must be less than 600 metres from a central active hub. Additional passive hubs, each accommodating up to three devices with cables up to 30 metres in length, can be provided at distances of up to 30 metres from an active hub.

Token Ring

This is similar to Token Bus, except that one station acts as a monitor station, providing tokens that allow other stations to send data. The other stations then wait to see when they can become the monitor. Each station has a 32-bit address; the first two bytes for the LAN ring number, the rest for the station number. Ideally this address should match the Internet Protocol (IP) address used in any Network Layer protocol, ensuring that each station has a unique worldwide identification.

Token Ring networks run at 1, 4 or 16 Mbit/s, with some types of Token Ring card accepting dual-speed operation. This benefit is easily lost, since the whole system slows down to the rate of the slowest card. Each station can be connected to a Type 1 STP cable via a wall plate and an adaptor cable or to a Type 3 UTP cable via an UTP transceiver.

At 4 Mbit/s a ring can cover up to 385 metres with Type 1 cable or 145 metres with Type 3 cable. A typical multi-station access unit (MAU) can connect up to 8 stations. The MAUs at different sites must be linked with Type 1 cable. A network made up of Type 1 cable can have up to 260 nodes with 33 MAUs. With Type 3 cable this falls to 72 nodes and 9 MAUs.

Wide Area Networks

A wide area network (WAN) is similar to a LAN but works over a larger area, although the dividing line between the two isn’t accurately defined. A WAN is often used to join together several LANs, each located at a different site.

For intermittent connections a dial-up telephone line and modem with a LAN modem server is adequate. More intensive links require a dedicated circuit, either in the form of an analogue twisted-pair circuit and modem or all (or part) of a digital trunk route. The latter, operated by a communications authority (PTT), can be used to create a private data network (PDN). Such links often use packet switching (PS) or asynchronous transfer mode (ATM) techniques.

The following systems can be encountered, although not all these are true networks:-

Integrated Services Digital Network (ISDN)

This is the most common PS trunk route. In Europe the coaxial cable or fibre-optic version runs at 2.048 Mbit/s and provides up to 30 B-channels at 64 kbit/s each, plus extra D-channels at 16 kbit/s. The ISDN2 service provides two B-channels via a twisted-pair cable, effectively giving 128 kbit/s, plus one D-channel. ISDN4 uses two twisted-pair cables to double this rate.

These services can be connected to any Mac OS computer via an appropriate PCI card or interface. Other networks, usually in the form of Ethernet, can be connected to ISDN via a suitable router.

Asymmetric Digital Subscriber Line (ADSL)

ADSL uses special electronics at the local telephone exchange to provide an astonishing data rate over a conventional telephone cable. In theory, it offers 9,216 kbit/s, that’s around 1.1 MB/s, for data that your receiving ‘downstream’ and 640 kbit/s for what you send ‘upstream’. In reality, it may only be 384 kbit/s to 2.5 MB/s downstream and 128 to 256 kbit/s upstream. Even so, this is much faster than a 56 kbit/s modem and more than adequate for conveying multimedia data.

Some cable television companies offer alternatives to ADSL, such as blueyonder from Telewest.

Asynchronous Transfer Mode (ATM)

ATM supports multimedia voice and video information for desktop conferencing and high-resolution video broadcasting, running at up to 622 Mbit/s. It combines PS with the traditional circuit switching techniques used in a telephone exchange. To use ATM you’ll need suitable hardware, usually in the form of a PCI ATM card that you can install in your computer.

Fibre Data Distributed Interface (FDDI)

FDDI runs at 100 Mbit/s using duplex fibre-optic cable in a ring configuration. It can support 500 nodes over a distance of 100 km, with up to 2 km between active nodes. It’s often used as a backbone or spine PDN that links together several LANs. Since it uses fibre-optics it doesn’t suffer from interference caused by differences in electrical potential between buildings.

An FDDI concentrator can be used to create a partially star-like configuration. In critical systems the duplex cable can form a complete loop via all the devices and back again, effectively creating a primary ring and a secondary ring, the latter maintaining operation should the primary ring fail.

To use FDDI you’ll need suitable hardware, usually in the form of a PCI FDDI card that you can install in your computer.

Metropolitan Area Network (MAN)

A MAN is a dual-ring system that runs at 155 Mbit/s. It’s operated by the PTT and can be shared by several organisations. Data flows in opposite directions in each ring and every device is given a time slot in which it’s permitted to send data.


A rather dated public network for linking LANs at 64 kbit/s, falling to 8 kbit/s at peak periods. Devices designed for X.25 operation only a require a modem to connect to this kind of system. Other networks and devices must be connected via a packet assembler disassembler (PAD), plus a modem if there’s not one in the PAD itself. The X.28 standard specifies how a PAD should operate when used in conjunction with a device fitted with an RS-232C port.

The X.25 specification itself defines the Physical Layer as X.21, the Data Link Layer as Link Access Protocol B (LAPB), a simplified form of HDLC, as well as the Network Layer. The latter makes connections using a virtual call or permanent virtual circuit. End-to-end checking is accomplished in the Transport layer.

Frame Relay

Similar to X.25, but running at any speed between 64 kbit/s and 2.048 Mbit/s with end-to-end checking in the Data Link Layer. It uses Link Access Protocol F (LAPF), derived from the LAPD system in ISDN, and provides permanent virtual circuits.


MacWorld magazine (UK), IDG Communications, 2003-2004

©Ray White 2004.