I will attempt here to give the reader an overview of the different pieces of work that are
included in this book.
Computer communications, of course, starts with the communication of a "bit" of information.
A bit is 0 or 1, and all kinds of information can be coded into sequences of bits. However, any
form of communication between a transmitter and a receiver will involve some physical media and
a bit itself has to be made measurable to the receiver. There have been, over the years, many
ways of doing this and they share certain principles that are used to increase the efficiency of
communication. This is the subject of the first paper, authored by Pierre Humblet. The
transformation of bits into signals that can be detected by physical devices involves both analog
and digital coding. One often is interested in the maximum rate at which over a given channel
such coders and decoders can be found. This has been a vital area in communication and
information theory, and today the modem in the personal computer is, in fact, one of the earlier
devices used to transmit digital information, meaning zeros and ones.
When there is a multiplicity of receivers and transmitters, digital information may be queued
at various sites because of variability in the generation of messages. These queues create
congestion in the network linking the elements. This can be modelled as a flow of messages
between nodes of a network where capacities of the links and the speed with which a message may
be switched from one link to another determines the performance of these networks. If one
attempts to model in great detail the time required for messages to travel on a link and get
switched onto another, the analysis of delays experienced by messages in the network becomes
virtually intractable. However, by modelling the times by random variables and making certain
simplifying assumptions on them, one can get some useful formulae about queue sizes and delays
experienced by messages. The principal tools used in this analysis are described in the paper by
F. P. Kelly. This paper shows clearly how certain assumptions simplify the analysis and yet yield
results which give insight into the behavior of real systems. On the other hand, it also shows how
small changes and assumptions make mathematical problems considerably more challenging.
Although making simplifying assumptions may be alright when investigating the overall
behavior of a computer communications system, the detailed behavior is critical to verifying the
logical properties of the system. By "logical properties" we mean the precise description of exactly