Title: The Wall

Dear Francis,

After having read the book reviews in the last C Vu, I thought I should start a discussion about them. I am concerned about the quality of the reviews. From quite a lot of them it was clear that the reviewers had very little knowledge of the topic the book was about. Some reviewers stated it explicitly, while from other reviews you could only notice it if you knew the topic and/or the book.

Some of the reviews reminded me of a course on advanced C++ I presented once. The participants had been in a previous entry-level course and had been really happy with it. But during my course it turned out that they had at least partly learned completely wrong things in that course.

The problem here was the same that most reviewers have: as long as you learn about something, you believe what your teacher or your book tell you. And so you are not in a position to comment about it. At least not about the technical contents. What you can do is to comment about the way and style of presentation. But that is only secondary. The primary criteria to rate a book must be its technical accuracy.

I believe that these problems are an inherent consequence of the current review procedure: reviewers choose a book they want to review, pay a handling fee to receive it and owe a review about that book. Of course, most people will do that only if they see some benefit in it. And this will in most cases be that they learn something new. So most reviews are written without the necessary knowledge to correctly rate the book.

So, what could be done to change that? One idea is to let reviewers write about books they already had (and used) for some time and where they really feel comfortable with the topic of it. And in return for the review, they get a new book for which they do not owe a review. But of course they are invited to write one about it as soon as they really feel qualified for it.

This is just a thought, but I think this would greatly improve the quality of our reviews, which is currently quite low in parts.

Actually, I share your concern. However, there is a problem with getting reviews out in a timely fashion. I think that the overall standard of our reviews is much better than those I see elsewhere. This is not to be complacent, I welcome further improvement. Technical accuracy in areas relevant to the target reader is very important. I do not get very concerned with side issues. The fact that many of the code examples in Design Patterns are poor C++ code does not worry me because readers of that book should know better and are more interested in the main thrust of the book. K&R (The C Programming Language) would be better as a book for novices if it not have an implicit expectation that the reader will be using a Unix based system.

Ideally, I would like to see books reviewed twice, once by a target reader, and once by a domain expert. As a voluntary organisation, ACCU just does not have the resources to ensure this happens. I would, however, hope that domain experts would add a second review whenever possible. Even then we have a problem, what qualifications make someone a domain expert?

I would welcome the opinions of other readers.