Review of Java Book by Davies

This book is accurately described by the author. Chapters 1 to 5 are a conventional 
discussion of Java that omits (deliberately) the powerful user interface an graphics 
capabilities. Chapters 6 and 7 describe two special features of special relevance (Chapter 
6 - visualization and chapter 7 - numerical libraries. Chapter 8 is embryonic but must be 
expanded if book published as it has the meat of science specific Java considerations.
	I teach the subject (Java for Science) and would not use the current book for 
several reasons described below.
	However some Java books sell in large quantities and this book is quite well 
written and "simple" (the cynic would say simple-minded). Thus I believe the book will 
be accessible to a large audience. Thus it could be a viable business proposition for 
Wiley. I expect it will have a relatively short life as others will write much better books
	Now I give some comments:
1) There are essentially no serious science programs in the book. The libraries of chapter 
7 are serious but we only discuss calling sequences and not the Java language and 
expression issues. Chapter 8 has some beginnings in this area but it is not a strong 
chapter at moment. Note that others have produced interesting codes, which are 
available on the Web.
2) The book does not consider issues relevant to programmers with a Fortran 
background (most important community?). The author appears to be a C++ expert.
3) The author seems unaware of work in area. There are many important points omitted 
in the book but published (on web) in recent conferences on subject of scientific 
programming in Java. The last one (June 21 ACM PPoPP) is particularly relevant.
4) The author does not discuss significantly some key features of using Java for good 
scientific code. Efficiency suggests that one use simple types and not objects (elegant 
those these be). Small methods and overloading (common in Java) are difficult for 
compiler.
5) I think the software of chapter 7 could be interesting as it is nontrivial. However it is 
not a "community standard" and this area/software will surely change rapidly as 
public domain packages such as Lapack get converted to Java.
6) The chapter 6 software appears inferior to public domain packages available in the 
U.S. over the web.
7) The book is as I said, clearly written and so even with limitations could attract 
readers.

Now I give a few strategic points:
8) There are several wonderful general Java books which are well written and come with 
a rich set of web based resources such as example applets, tutorials etc. The best of 
these books are now in second editions and have been evolved through extensive use. 
This Java for Science book (and indeed any mere mortal) cannot compete with these 
books which have large teams and substantial fiscal investment. Thus I think you 
should build on top of the commodity Java books. A Java for science book should 
choose one or more of these books and focus on adding values in those critical areas 
(chapters 6-8 of current draft) where the basic books are deficient.
9) I disagree that is a good idea to avoid the graphics and networking ideas in Java. Most 
science problems do a lot of I/O and it is the convenient integration of numerics, 
graphics and networking that will be attractive to many scientific programmers.
10)  The conferences (mentioned in 3) above) point out many other features of Java in 
areas of parallelism, co-ordination etc. This again emphasizes difficulty with 
"stripped-down" Java approach in this book.