So I first note that I often pass your Cambridge University Press offices by as I 
come by train from London when I visit my parents who still live in Cambridge. I was an 
Undergraduate and Graduate student there. So now to your questions:
I would agree with Davies that Java is likely to be a significant scientific 
language. Further I find the book rather well written. However I have some reservations 
that I detail below.
1) In some sense the book is premature as some key issues about the language and how 
to use it are a subject of active debate between Sun and the Scientific Computing 
Community. As an illustration of this, note the conferences: 
http://www.acl.lanl.gov/iscope/, http://www.cs.ucsb.edu/conferences/java98/, 
http://www.npac.syr.edu/users/gcf/03/javaforcse/acmspecissue/latestpapers.html, 
http://www.supercomp.org/sc97/bofs.html . These debates could either change 
language/Virtual Machine and/or define coding practices for constructs such as 
complex.
2) It is encouraging that Visual Numerics is using Java and JSGL seems a reasonable 
graphics library. But this is a rapidly changing field and already very powerful Java 
graphics libraries are becoming available in the U.S.A. These are tracking the new 
Java3D standard and being developed by projects such as the NSF supercomputer 
centers. Inevitably these will be preferred to JSGL. Again Linpack is outdated 
technology and many other Java numerical mathematics libraries such as those 
produced from LAPACK will challenge VNI.
3) I judge the scientific computing specifics of this book as rather rudimentary and 
ephemeral. The Java tutorial is well written but incomplete. I would not recommend 
for a course aimed at teaching Java for Scientific Programmers. Rather I would 
recommend one of the many excellent basic Java texts and pick and choose topics. I 
would as a lecturer, enhance the book by additional material such as that in chapter 8 
of proposed book. There is not much of this yet (tutorial Java scientific codes) but 
there isn't much of chapter 8! The freeware world will soon produce a wealth of such 
examples and although incoherent, this is useable as add-on material to a basic text.

So I view area as quite promising and the author as talented with a good pedagogical 
style. However this particular book has the difficulties outlined above.