A great expositor of mathematics, Paul Halmos, gave some advice on this question.
Below is a slightly modified version of his advice.
A good book review is a chatty expository essay on a currently interesting subject. The purpose of the review is to provide in at most two pages a brief description of what the book does (or could have done) in three or four hundred.
A book is an excuse for a review. A dangerous demagogue is as good an excuse for an essay on politics as a wise statesman; it would be wrong to refuse to write about something because the last news about it was bad. Similarly, what is important about a book review is that it is an essay on an important subject; it is wrong to refuse to write about it because the book is "bad" and does not "deserve" a review.
The review should tell its readers where the subject of the book comes from, where it is now, and where it is going. What are its connections with other, perhaps better known, subjects? What should most computer scientists know about it? What makes it interesting? Why should a book have been written about it now?
An annotated table of contents is a bad book review.
A book review preoccupied with value judgements is also bad. Lavish praise is usually dull, and nasty sarcasm usually reveals more about the reviewer than about the book.
Facts about the book can, of course, be interesting. If, for instance, the book does not have an index, the reviewer might want to mention that. The reviewer might also want to mention that the book has an unusually helpful list of notations, or that it is not organized into chapters in the usual way, or that it is full of misprints, or that most of it is a verbatim reproduction of earlier papers by the author (or by someone else).
As for judgements the reviewer may, of course, say (or imply) what he thinks. If he thinks that the treatment is bad, or the material is out of date, or the references are inaccurate, he is certain1y allowed to say so - but, on balance, the review should be the reviewer's exposition rather than a critique of the author's.
It is our desire to publish reviews of as many relevant books by Australian authors as possible. To do this we need not only a large number of reviewers, but also the co-operation of Australian authors and of publishers.
Please note - we gratefully receive all relevant books and software sent to us for review purposes; such materials become the property of the Australian Computer Society Inc. or the reviewer, irrespective of whether they are reviewed. These should be sent to the Editor-in-Chief.
It is important to note that copyright for all reviews received by the Journal is retained by the Australian Computer Society Inc.