Thanks for your kind words about my humble efforts, while that model
seems to have come up, let me say a few things about it, and about how I view
some of the threads being discussed right now. (I prefer to lurk, but when
my name comes up, I'm more than happy to contribute :-)
The model was done in showcase and jot. I have no expensive 3-d modeling
software, and unless it becomes MUCH less expensive, I will never buy any.
Showcase is a pretty good 3-d modeler as long as you stick to primitives.
However the most important modeling tool I have at my disposal is a good
text editor! Despite what people say they are working on, there are no 3-D
modelers that can both import and export VRML models right now.
The reason for the spectacular development of the web is that the
format of the files was completely open and human-readable. This meant that
anyone anywhere could use existing text-editing tools to quickly create large
amounts of content, sure editors that were specifically aimed at the web came
along eventually, and perhaps, in the future, most HTML authoring will take
place on tools like these, but what caused the GROWTH of the web was the fact
that ANYONE can do it. If we really want VRML to catch on like HTML did, the
most important factor is that people not be scared of it.
For this reason more than any, I think the entire thread of the
binary format is headed in the wrong direction, sure files need to be small
to transfer accross the net, but shouldn't that be up to the modeler?
I have discovered in my modeling experiments in VRML, that many 3-D
modeling programs create terribly inefficient models, up until now, this did
not really matter, very few real-time 3-D systems existed, and those that did
either had enough graphics horsepower to just deal with it, of had
sophisticated software capable of fixing the inefficiencies at parse-time.
This may have been because the programmers figured that no-one would see the
actual scene anyway, and software doesn't know the difference. Well those
days are gone now (thank goodness) and modeler writers will have to be
concerned with how efficient their models are, not just how pretty they look
when you ray-trace them. Rather than talking about how to compress
inefficient files to make them quicker to transmit, why not talk about how to
make these models more efficient, this will make them quicker, both to load
and to parse (and render if your browser isn't very smart).
While hand-editing VRML documents may seem inefficient, it is
actually not that bad. Cut-and-paste comes in handy as does the "delete"
function. Almost every model I have had to hack was at least twice as big as
it needed to be. Redundant Groups and Separators around individual objects,
normal definitions for objects that didn't need them, (use creaseAngle
correctly and you will find that VERY few objects really need to have normals
specified), redundant vertices in indexedFaceSet objects (a cubeoid can be
described with 8 vertices, many of modelers describe it using 36 vertices,
one for each corner of each triangle). These are issues that modelers and
whoever writes format-conversion programs should be worrying about not this
group, you've defined a format, now let's see how people use it before we go
off and re-design it. (One idea that has some merit however is shortening
the keyword names, with all due respect to the Inventor team, VRML is NOT
Inventor, why are we clinging to this long-winded sintax? Why use "rotation"
when "rot" is so much shorter?)
That's enough ranting for now, I'm sure that I've opened myself up for all
kinds of flames, so I will see what kind of responses I get before posting on
any of the other subjects I have opinions on :)
Robin ("Does an asbestos suit cause skin cancer?") Hayes
-rdh
Robin Hayes, Silicon Graphics Inc. (415) 390-2456