Ethernet - the current favourite

The term "Ethernet" describes a family of 10Mb/s networks that rely on "collision sensing" to allow many users to connect to a data highway. We won't be discussing Thick Ethernet: thin ethernet (also known as Cheapernet or 10base2) uses thin 50ohm co-axial cable, daisy-chained between stations, whereas Twisted Pair (TP) Ethernet, also known as 10baseT, uses twisted copper pairs, such as our structured wiring, with one circuit (and one hub port) per station.

Ethernet is designed to carry many different networking "protocols" over the same network. Protocols can be divided very roughly into 'local area network' (LAN) protocols, that are intended to enhance the functionality of groups of stations that are located relatively close to each other (e.g within the Department or at least within the campus), and Wide Area Network protocols, that are designed to allow communication between computers that are located at a distance from each other - maybe on the other side of the world. The distinction is by no means hard and fast, as it is common to use some WAN protocols (e.g TCP/IP) between machines located in the same room; LAN protocols can sometimes be used between distant machines, although at a considerable loss of performance.

As examples of LAN protocols we may take Novell and Appletalk, whose purpose in practice is often to allow computers (PCs or Macs) to seamlessly access data, or to access devices such as printers, that are located on other machines nearby. In this way it is not necessary to duplicate data across machines, or to supply every user with their own printer. Machines that are coupled by LAN protocols often have the same management, and can therefore in a sense 'trust' each other.

Our prime example of WAN protocols would be the TCP/IP family, as used on The Internet. These are designed to give access to distant computers which, in general, are under different management and must be considered potentially hostile. Traditionally, access to such machines is not seamless, but requires the user to invoke a specific application (such as TELNET, FTP etc.) and to identify themselves to the remote computer before any service is possible.

Here then are some protocol families set out with brief comments.

Internet Protocol (IP, also known as TCP/IP)
This protocol is likely to be important for users who have any interest in the 'outside world'. It allows you to communicate via the campus IP network and JANET IP service (JIPS) with the whole world-wide "Internet".
Used primarily between VAX/VMS systems. This protocol is also capable of supporting world-wide communications.
Used primarily between Macs and LaserWriters. Communications are normally limited to a single campus: wider scope is possible, with care, but is not usually attempted in our environment.
This was meant to be the future protocol for JANET. However, IP protocol seems to have won, for the time being at least.
LAT (Local Area Terminal)
Used only for terminal access (from terminal concentrators) to computers on the same network.
A proprietary LAN protocol. This will be important to those wishing to access licensed PC software on the Departmental applications server.
And many more.
To get some concept of the use of multiple protocols on the same network, it may be useful to think of the analogy of a room full of people, each of whom speaks one or two languages out of many. When someone speaks, only the people who understand that language will react: the others just disregard it (but nevertheless they must avoid speaking when someone else is speaking, regardless of whether they can understand what is being said).

The word "network" is rather vague, and can be understood in various different ways according to context. As far as we are concerned, an Ethernet "network" refers to all the pieces of Ethernet that are connected together by repeaters or bridges, and have the same IP subnetwork number: the pieces are not necessarily all of the same kind, but could be partly TP Ethernet, partly Cheapernet, partly Fibre-optic etc. as we will see. For convenience, we usually refer to a network by its IP subnetwork number. Networks are joined together by means of "routers" to form "an internet", and the whole assembly of networks that are joined together world-wide is known as "The Internet".

There are in fact several Ethernets within our Department, linked together by routers: the following diagrams may be of interest, although not necessarily up-to-date.

Overview of the Ethernets in the Department

Our connection to the rest of the campus (and hence to the rest of the World) is via a rack of equipment in the basement, that connects to multi-fibre cables laid around the campus. This equipment (in the dashed lines) is operated by the Computing Service, and offers us (presently) two connections, one a bridged connection to a campus-wide Ethernet (IP subnet 4), and the other a Departmental ethernet (subnet 45) that is routed to the campus 100Mb/s FDDI ring. Most users are connected to the latter.

The theorists operate their own private network, number 202. Instead of having a dedicated router, they have one computer with two Ethernet interfaces, connected respectively to network 202 and to (as it happens) the campus-wide Ethernet, IP subnet 4. This computer runs routing software in the background.

The student PC cluster also has its own IP subnet, primarily in order to shield the rest of the network from possible blunders or malicious interference. Here, the routing function is done by a dedicated PC.