In this paper, the author tries to describe the way a digital computer works: from gates, through instructions, data encoding, and interrupts, to peripherals of more than a dozen types. He tries to do this in 29 pages plus a nine-page appendix on semiconductor technology. He tries to do it for an audience described in the book’s foreword as
(1) physicians who were educated in the pre-computer era,
(2) medical students and house staff being educated now,
(3) computer professionals interested in medical applications, and
(4) readers outside of the medical profession and the computer science community who are interested in the role of automation in health care.
This is a tall order. The paper comes impressively close to filling it. In clearly-written prose, Spangler takes the reader through the overall structure of a computer; its major components; information, binary numbers (including floating point), and character codes; word length issues; gates, combinatorial circuits and flip-flops; instruction sets, including subroutines and interrupts; and peripheral devices, with emphasis on mass storage. A necessary byproduct of covering all this material in such a short space is a condensed, economical writing style. Statements are made and concepts are described with little explanation or follow-up. All the elements of a system are covered, but the pieces are not put together to create an overall sense of how the whole actually works.
This condensation makes the material rough going for three of the four groups in the book’s intended audience. (Computer professionals will skip this paper.) The material will repay careful study with an understanding of how the parts of a computer work. But this means careful study. Virtually every sentence introduces a new, complex, and non-intuitive concept that will not be revisited but is assumed to be understood thereafter. A knowledgeable expert is almost mandatory for explaining sentences such as “Immediate signifies data contained within the instruction,” presented as a note to an instruction list with no further explanation or example. Flip-flops are explained in 17 lines plus a circuit diagram. Floating-point notation gets half a page with no examples. My preference would have been to leave out the digital logic and the appendix on semiconductors and use the space to explain more and connect more.
As a computer architect, I found the material almost totally free of technical errors. But there are a few. The concept of firmware does not depend on PROM technology. A book with a 1986 copyright should not refer to 2400-baud modems as having “recently become available” or “probably approaching the practical limit for common telephone line transmission.” Indexed and based addressing are separate, albeit related, concepts. But these are nit-picking quibbles. Taken as a whole, the material will not mislead its readers.
To sum up, there is a lot of content in this paper. I fear there is too much for most of its intended readers. It is like trying to drink from a firehose. The water is there--but the poor, thirsty drinker is overwhelmed.
Finally, I would like to express my gratitude to Paul M. Hardy, M.D., and Arnold F. Mazur, M.D., for providing physicians’ reactions to this paper. The responsibility for interpreting (or misinterpreting) their comments is, of course, mine.
