Title: Letters to the Editor

The proposed change to the book reviews was enough for this email.

On the book reviews/ratings etc. discussion, I think that the base issue is what a book rating is for. One of the main reasons for such a rating is for someone to chose which book might be a good investment for some particular purpose (e.g. learning, reference etc).

If we give ratings on books (i.e. just a number or a conclusion separately from the full review) it would generally be used so that people can find the excellent books quickly. (Why would you want to know whether a book is average, or really, really bad? You should probably avoid it anyway. If you inherit it, you might want to read the entire review to find out what is right/wrong with it.). Perhaps we would have to qualify what the review rating is designed to be used for.

The meaning of 'excellent' is probably going to be different depending on who you are (super-expert/beginner) and quite possibly what you are going to do with it (games programming v. financial applications v satellite control. Reference, or discussion of finer points of syntax? etc.). Also, what one would consider excellent would be expected to change with time. (What would the original K&R book on C rate as 15 years ago? And what today?)

I think that rather than attempting to rate all books, an ACCU rating of books that we would consider indispensable might be useful. This could represent a general consensus of the membership, rather than just a single reviewer, or even a review panel. (As lots of people would have read Stroustrup, Meyers and a host of other top-rated books it would not involve a huge amount of postage or even necessarily of organisation). It would be then be reasonable to review this list once a year, to see if there are any missing or ones that should be removed from this list. The number of times books crop up in references in Overload might be an interesting place to start.

How many books should be on the list? Possibly only 10 core C++ ones - and another 5 or 10 for specialist purposes (and an appropriate number for other languages)

It would also mean that we might be able to supply different people's opinions and any caveats on the books - which would be interesting reading in its own right.

There are almost certainly problems with this scheme. Perhaps there are other reasons people like to have ratings. Perhaps this could be just an adjunct to the existing book review rating scheme (recommended, highly recommended etc).

James Roberts <James.Roberts@logicacmg.com>

Thanks for such a great email which more or less reflects what I've been saying for quite a while!

The point over what constitutes the ratings is something which does have to be ironed out.

As you're aware, we have 4 ratings; not recommended, nothing, recommended and highly recommended, with recommended being like a grade 2 degree (2i or 2ii - recommended or recommended with reservations). There is nothing to say what has to done to achieve one of these grades.

What has been proposed is that the reviewers have a set of criteria to judge the books against. It is not a tick list as it still allows for the reviewer to use his/her judgement - I have reviewed some books which while technically correct, have been written so badly that their use is very limited. A simple tick system would have gained it (say) a recommended, but the judgement would drop it down. In lay terms:

Highly recommended : It's been written by Stroustrup, or Josuttis

going down to

Put it back on the shelf or if you've bought it, demand a refund : anything in the "for dummies" series, Schildt or "C++ in 21 days" type books.

Of course, the review system is still in the early days domain, so what will happens is still to be determined.