Indispensable.
As a Computer Science PhD candidate, I was absolutely shocked to see a book like this even exists - it seems so specific, but writing in the Computer Science field is often terrible, and Justin Zobel's book painlessly dissects what makes CS papers so bad, and how to avoid those problems.
The book covers everything, ranging from how to conduct research to how to perform experiments, and, of course, how to actually write the paper. No stone is left unturned, Zobel doles out advice for grammar as well as figure and table design, software to use, bibliographies, and how to best structure a paper to effectively communicate ideas. He even includes sections on ethics, giving presentations, and refereeing papers.
Most of the book consists of an explanation of how to write effectively in Computer Science, followed by one or more examples of how to fail in that specific way, followed by an example of how to succeed. These comparisons make it very easy to understand Zobel's points, so the book is very easy to digest and understand.
If you're in Computer Science, and you're writing papers with the goal of publishing, this is an absolute must-read, end of story.
The only bad thing I can say about this book is, now that I know what kind of mistakes to avoid in CS writing, I see them everywhere. Clearly, more people need to read this book.