Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!usc!apple!well!levine
From: levine@well.sf.ca.us (Ron Levine)
Newsgroups: comp.graphics
Subject: Re: Practical Intro to PHIGS (new book)
Summary: Not insufficient detail, just poor explanation
	 .
Message-ID: <24373@well.sf.ca.us>
Date: 23 Apr 91 09:18:41 GMT
References: <1991Apr19.175910.2922@Stardent.COM>
Lines: 94



I am befuddled by Jan Hardenbergh's insistence on calling this 
book "great", even though he recognizes many of its shortcomings.  
I am also befuddled by his refusal to acknowledge many other of 
its flaws.

>I stand by my characterization that this book is great.  I've had
>another few days to look at it, too.  I do concur with Ron Levine in
>that the does not go into lots of detail in many places.  But it is a
>339 page book - including lots of appendices and examples.

I did not complain about lack of detail so much as the fact that 
the quality of explanation is frequently poor.  Since Jan 
focusses on my complaint about the definition of "implicit 
regeneration", let us look at the complete paragraph. 

   "Many applications build and modify models incrementally and 
   iteratively.  Not all workstations, however, have the ability 
   to modify parts of a displayed picture _selectively_.  On some 
   workstations, such as a laser printer, the only way to bring 
   the picture up to date is to redraw it completely.  Redrawing 
   the entire picture in this way is known as _implicit 
   regeneration_.  It is implicit because the redrawing is not 
   explicitly requested by the application, which specifies only 
   the incremental changes required to the picture.  Implicit 
   regenerations occur under the control of PHIGS." 

Now the first sentence creates the wrong impression, because, in 
fact ALL applications build and modify models incrementally, in 
the sense that they can call only one PHIGS function at a time.  
The question is whether the picture should be regenerated after 
each PHIGS function call which redefines the picture, or only 
after a batch of such calls.   

The second sentence overlooks the fact, that in general, NO 
workstations have the ability to modify parts of a displayed 
picture selectively, at least not correctly.  On a CRT, you can 
delete a primitive by painting it out in the background color, 
but in general you will not produce a correct picture because you 
will damage whatever primitives lie under the deleted one.  
   
Thus, the "laser printer" in the third sentence is a confounding
red herring.  The main clause of this sentence is essentially 
true of ALL workstations.   The key is understanding that "redraw 
completely" = "regenerate" = "traverse posted structures".  It 
means traversing all the posted structures, respecting the 
relative priority of the root structures, the order of primitives 
within a structure, and whatever hidden line or hidden surface 
methods are in effect.  That is the only way to get a guaranteed 
correct picture, no matter what the output device. 

Now we see that the fourth sentence, which purports to define 
"implicit regeneration", is meaningless because of the fuzziness 
of "in this way", which seems to have something to do with laser 
printers.   The fifth sentence would be true if we had a valid 
definition of "implicit regeneration" in the fourth, but we 
don't.  The last sentence is sort of true, but inadequate 
because here is where the important concept of deferral mode 
should be introduced.  

The last sentence should rather say something like:   "Implicit 
regeneration is regeneration automatically instigated by PHIGS in 
response to any picture changing function call if the deferral 
mode is ASAP, or if the deferral mode is BNIG or BNIL and an 
input interaction is underway.  If deferral mode is WAIT, then 
regeneration occurs only when the application explicity requests 
it by calling UPDATE WORKSTATION or REDRAW ALL STRUCTURES." 

Again, nowhere does the text mention the fact that on most 
workstations, most of the time, the only sensible deferral mode 
to use is WAIT.  If you want approximately correct incremental 
drawing on a CRT, then you might also use modification mode UQUM 
(quick update methods), which might or might not be supported on 
your workstation type in the implementation you're using.  It 
wouldn't make sense to use UQUM on a laser printer, but it 
wouldn't hurt either, provided that you remember to explicitly 
update the display at the end of the set of incremental changes. 

Now this is just one example.  There are many other cases of poor 
explanation, or explanations that miss the essential point.  I 
previously pointed out the lack of justification of the primitive 
aggregating feature of POLYLINE SET WITH DATA.  The definition of 
"right-handed coordinate system" in Sec. 3.1 is just ludicrous 
and has no practical value. 

I concur with Jan that this is the best PHIGS book out, because 
it is the only one.  As an introduction to pre-PHIGS PLUS PHIGS 
it is a lot better than some of the vendor documentation, but 
probably no better than some others.   But anyone who buys it on 
the strength of the title's claim that it is a "practical 
introduction" to PHIGS PLUS is being duped. 

Ron