The EXPRES project started with a grand goal: to allow automated submission of multimedia proposals to the National Science Foundation (NSF), potentially saving tremendous amounts of time and paper. NSF proposals are often written by several researchers at various sites who use different word processing software. Passing incompatibly formatted documents around is a problem. The authors approached this problem by defining a common intermediate form based on the International Standardization Organization’s office document architecture (ODA) and writing two-way translators between their ODA format and the Carnegie-Mellon Andrew Toolkit, BBN Diamond, and Interleaf, as well as a one-way translator from ODA to UNIX troff.

Any such translation has several conflicting goals: imaging fidelity, which preserves the exact appearance of the formatted document; content fidelity, which preserves the text and images within the document; structural fidelity, which preserves document features such as paragraphs, footnotes, and page headers; and editing fidelity, which lets users manipulate the translated document in the same way that they would have manipulated the original. Since the project goal was to process editable documents rather than final page images, content and structural fidelity were most important. Content fidelity was easy, since all of the document languages used represent text as strings of ASCII text. The authors did not handle any images other than bitmaps because translating all the various picture formats seemed too difficult given their three-year time frame. Even translating different bitmap densities was hard enough that they limited their bitmaps to 72 DPI, the lowest resolution in common use. The hardest problem was structural and editing fidelity. Source documents typically have styles so that, for example, a single section header style can be applied to all of the headings in the document. A satisfactory translation needs not only to apply the heading style to all of the headings, but to do so in such a way that the user can change all the heading styles together rather than one at a time. Also, in many cases, different document processing systems had similar but not identical constructs, which meant that a straightforward translation could produce occasional strange and undesirable results; for example, when one formatter automatically numbers footnotes and another does not, what does the translator do with footnote numbers?

The first third of the book is a narrative describing the authors’ experience in designing and writing the translators. ODA turned out to be a difficult intermediate form for several reasons: ODA is an enormous and cumbersome language, so that handling all of its options is tedious; it often provides many equivalent ways to specify something; and the ASN.1 file format is verbose and difficult to read, write, and debug. ODA is also lower level than the other languages used, so retaining structure and style information was hard. An 80-page chapter describes in stupefying detail how all of the formatting features the authors handled were mapped to their intermediate format, a mapping made extraordinarily complex by the limitations of ODA.

The only experience they report with their translators is some demonstrations to a few hundred people, along with a few screen shots of the first page of a translated document. It is not clear that the observers understood what was being demonstrated, since some of them decided that the project was a failure because it did not preserve perfect visual fidelity.

The last two thirds of the book are a reference manual for the ODA and raster graphics toolkits developed to support the translators, including a short discussion of their design and development methodology. The ODA toolkit appears to be helpful to deal with the morass of detail involved in the use of ODA. The raster toolkit may or may not be useful, since the authors admit that it was completed so late that they were not able to use it in any of the translators.

As a study of multimedia document translation, this book is unsatisfactory. It shows that translation is not trivial, but anyone who has tried to move a document from one word processor to another knows that. The authors’ results are so limited, and so affected by their difficulties dealing with ODA, that it is hard to tell whether translation is fundamentally as hard as they found it to be or whether different decisions early in the project would have made their job much easier.

The authors typeset the book using Scribe, a well-known document processor that they used for all the project documentation, but which EXPRES translators neither read nor write. The book suffers from insufficient editorial attention by the publisher. Much of the writing is murky and repetitive, and typographical errors obscure the meaning of some of the text and figures.