Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!pacbell.com!pacbell!well!levine
From: levine@well.sf.ca.us (Ron Levine)
Newsgroups: comp.graphics
Subject: Re: Practical Intro to PHIGS (new book)
Summary: Maybe welcome but not great
Message-ID: <24283@well.sf.ca.us>
Date: 18 Apr 91 01:36:53 GMT
References: <1991Apr16.022646.6268@Stardent.COM>
Lines: 182



Jan Hardenbergh reviews the new book on PHIGS and PHIGS PLUS on 
the basis of an all too cursory inspection:

>I just got a copy of "A Practical Introduction to PHIGS and PHIGS PLUS"
>by Toby Howard, Terry Hewitt, R.J. Hubbold and K.M. Wyrwas.  ISBN
>0-201-41641-7. I've been looking at it off and on all day now and am
>more and more impressed each time I pick it up.

Then he concludes by calling it a "great book".

I have had a copy of this book for about a month.  Like Jan, I 
was at first very impressed because the book is indeed very 
attractive.  But on a careful examination I find that it has many 
faults and falls far short of "greatness". 

To be sure, as the first matter to appear in an important vacuum, 
it will be helpful to many who are struggling to get started with 
PHIGS in the face of inadequate vendor documentation.  But some 
of its inaccuracies and oversights are sure to obfuscate and to 
hinder the progress of some readers, some of the time.  
Therefore, I feel compelled to counterbalance Jan's excessively 
enthusiastic endorsement with some negative comments and caveats 
that this book is not all that it claims to be.  The material on 
PHIGS PLUS, in particular, is inadequate and sometimes even 
erroneous.  But even the more elementary material suffers from 
poor explanations which frequently miss the point. 

One of the first objectionable things I noticed is that from the 
beginning there is an implicit assumption that the default 
display update deferral mode is ASAP, which may be true in some 
implementations, but not in some important ones such as 
DEC PHIGS and Domain/PHIGS.  The PHIGS standard does not specify 
a default deferral mode, but it is certainly true that on most 
interactive systems, for pictures with any complexity at all, 
deferral mode ASAP is a bad choice, except perhaps for debugging 
purposes.   Thus, a number of implementations use WAIT for the 
default deferral mode. 

As a consequence of this implicit assumption, many important 
unqualified statements in the first half of the book (e.g. 
"posting causes the structure to be traversed", and "traversal 
may be thought of as a continuous process") are false for 
implementations in which the default deferral mode is other than 
ASAP.   Moreover, all of the example programs will fail to produce 
the expected picture on such implementations.   Since display 
update deferral mode is not discussed until Chapter 8 (except for 
a mysterious hint on page 9), the beginning student using such an 
implementation is sure to be frustrated.  

Then, when deferral mode is finally discussed in Chapter 8 (and 
in the hint on page 9), the explanation largely misses the point.  
It seems to imply that you would want to use a deferral mode 
other than ASAP only on a hard copy device such as a pen plotter 
or laser printer, which is just not true.  In fact, interactive 
applications on devices such as CRTs are EXACTLY where you would 
want to use deferral mode WAIT, together with modification mode 
NIVE or UQUM.   Further, I think it is confusing and very bad to 
place the discussion of "dynamic modification accepted" in Sec. 
8.3, before the discussion of deferral mode and quick update.  
And, in my opinion, the definition of "implicit regeneration" in 
this section is just wrong.  (The PHIGS standard uses the term 
but never precisely defines it).   The whole discussion fails to 
mention the all-important point that traversal, or regeneration, 
is generally needed to be sure that the picture is correct,  no 
matter what the output device is. 

This observation of the difficulties with the book's treatment of 
display update caused me to examine many of the explanations 
closely, and in many cases I have found them wanting, especially 
in the PHIGS PLUS chapter.  For example, consider the description 
of POLYLINE SET WITH DATA:  

        "...it allows us to define a set of unconnected polylines 
        as a single primitive.  This can be useful for displaying 
        a variety of composite shapes constructed from individual 
        polylines".  
        
Of course, EXACTLY the same "variety of composite shapes" may be 
drawn using individual POLYLINE primitives.  The explanation 
ignores the primary rationale for aggregating a number of 
individual polylines into a single primitive: namely, that it may 
frequently result in important performance gains, by reducing the 
overhead of function calls, bus transactions, etc. 

I am particularly offended by the discussion of SET OF FILL AREA 
SETS WITH DATA:  

        "We will not go into more detail here; suffice it to say 
         that this degree of complexity is not something which 
        the everyday application programmer wishes to grapple with."  

I strongly demur.  While this primitive might not be the first 
that a PHIGS beginner would want to attempt to use, I certainly 
think that it is an extremely important tool in the box of the 
"everyday application programmer", and no book on PHIGS PLUS can 
be complete without a good discussion to help the reader with its 
complexity, as well as examples to demonstrate its power and its 
benefits. 

The discussion of the Phong reflectance model, in particular the 
significance of specular versus diffuse reflection, is far too 
terse in relation to what is needed by people who are not already 
experts on rendering--at least for the students I have had when 
teaching PHIGS in an industrial setting. 

The presentation of NURBS curves and surfaces is very bad indeed.  
It is non-illuminating, and contains erroneous statements as 
well.  For example: 

        "There are two types of B-spline: rational and non-
        rational.  Only rational B-splines have the property of 
        being closed under a perspective transformation, which is 
        why they are used in PHIGS PLUS".    

First, this ignores the important fact that "non-rational" B-
splines ARE rational B-splines in which all the weights are 1.  
("Non-rational" is a very unfortunate choice of nomenclature, 
evidently made by computer scientists.  Mathematicians would 
never use such illogical terminology.  Here we have a very clear 
example of the confusion that illogical terminology can cause.)  

Now, one of the most important properties of B-splines, rational 
or non-rational, is that they are invariant under affine 
transformations;  this is important because it means that the 
curve you get from a set of control points and weights does not 
depend on the choice of coordinate system.   The book ignores 
this property as well as several others--local control, convex 
hull, variation diminishing, etc.--which are all very important 
for explaining why B-splines, rational or non-rational, are used 
for the free-form curve and surface capability in PHIGS, IGES, 
and other systems.        

On the other hand, the one cited property, closure under 
perspective transformation, is not very important at all.  In 
practice, it really does not matter that the perspective 
projection of a non-rational B-spline may be a rational B-spline.   
Indeed, many authors on NURBS recommend that application 
developers use non-rational NURBS except when non-constant 
weights are absolutely required.  When is this?  The answer is 
that rational B-splines are needed for exactly representing conic 
sections (circles, ellipses, parabolas, hyperbolas), and quadric 
surfaces (including spheres, cylinders, cones, etc.) as splines.  
This is, I think, the REAL reason that the complexity of weights 
and rational B-splines has come into computer graphics. 

Further serious faults with the NURBS discussion: absence of any 
clue as to the geometric significance of the weights, and lack of 
any guidance on appropriate choice for the order.  Finally the 
discussion peters out with a weak reference to the books by Farin 
and the Killer Bees, which both are much too mathematically 
difficult for the supposed average reader of this book on PHIGS 
PLUS. 

It is not at all easy to write a good book on PHIGS PLUS.  I know 
because I have written one, in the form of a tutorial manual for 
internal use by a large corporation.   I spent an inordinate 
amount of time per finished page on the discussion of NURBS, 
trying to explain it clearly at about the right level for the 
non-mathematician reader.   So I hope it is understandable that 
my hackles are raised when I see such an incomplete job rushed 
into print to take advantage of a market opportunity, and then 
receive accolades based on an all too cursory reading from an 
expert like Jan Hardenbergh. 

The book is not without value.  Many of the nice features cited 
by Jan are indeed laudable.  But it is certainly not a great 
book, it has serious flaws, and it is not at all adequate as an 
introduction to PHIGS PLUS.  The authors may plead that PHIGS 
PLUS has been a moving target during the period of writing.   But 
then it is misleading to claim that this alleged "introduction" 
to it is at all "practical".  

I don't know anything about the other forthcoming books that Jan 
mentions, but I rather suspect that a truly valid and helpful 
book on PHIGS PLUS is still about a year away. 

        Ron Levine
        Dorian Research, Inc.
        (415)-535-1350