The Ethernet created for bandwidth explained itself as an ecological niche into which it fit perfectly; its advantages and limitations closely matched what the community could use. The contention principle, on which it was based, appealed to people's imagination and the bandwidth was more than adequate for the time. Limitation to a relatively small geographical area was not seriously felt.

The overwhelming attraction of the early Ethernets and rings was their simplicity. They were not networks in the true sense although the term Local Area Network (LAN) was adopted early. They were one-levels interconnect systems, with flat addressing and no routing problems.

As soon as Ethernets (or rings) are interconnected to give a genuine network, simplicity vanishes and all, or nearly all, of the problems associated with wide area networks come rushing in. This is particularly true if the resulting LAN is multiply connected. Nevertheless, methods of interconnecting Ethernets are available, and while not without problems, are perfectly adequate for present day requirements.

LANs will inevitably move to wider bandwidths, but it is not entirely clear how this will happen. Ethernets have already reached their potential. Rings operating in the range 50-100 Mbits/sec have been in use for some time and a standard (FDDI) exist. We are beginning to see experimental versions of rings operating in the gigabit/sec range.

The use of fiber optic cables as the transmission medium enables rings to cover a wide geographical area--well beyond a local area as that term is usually understood. However, large rings are unwieldy and packets take a long time to go round. Rings are perhaps best adapted for use in a backbone interconnect role. There is a growing interest in LANs based on VLSI cross point switches. At one time it had seemed that VLSI was not well adapted to the fabrication of cross point switches, because of the large number of pins required. This problem is now in the past and at least one 64 channel cross point switch is available.

The decision as to what should be centrally located, and what should be located in the user's office, depends on the speed of the LAN. Opinion on this matter has swung from side to side as users have weighed the advantages of independence against the need to have access to central services. As LANs move to higher speeds, it will become possible, if so desired, to separate the display processing from the display itself, and there will be no need for the user to have more than the bare display hardware on his desk.

There is an obvious advantage in locating equipment that is noisy, gives off heat and requires maintenance in a central computer room. Installing workstations in an office building can lead to expenditure on improved air-conditioning that can exceed the cost of the workstations themselves.