Mark - I've been doing extensive testing with Walkthrough Pro
on the Mac and as many non-PC convertors as I can get my hands on!
Here's what I've found:
Using Virtus Walkthrough Pro 2.0 or 2.0.1 (68k version only -
the PowerPC versions do not support DXF export, which is a shame). I use
the following options (Export->DXF 3D):
* 3DFACE (POLYLINE makes the model look like a wireframe)
* Color by object
* UNIX Line Termination
* Use Extrusions unchecked
I FTP this in binary mode to an SGI box and run dxftoiv on it.
In the end my file is bigger than the Walkthrough Pro file by at least
300% - some Walkthrough Pro files I've made are 1.5MB and they end up as
5MB to 10MB DXF files, and I sometimes have run out of memory during
translation.
Checking things with ivview, I find that often many of the polygons
end up being transparent looking at them from one side. So I install
gview from the Inventor Developer's Toolkit (From some IRIX 5.3 CD),
that's an Inventor graph viewer, grab my copy of Inventor Mentor, and try
to edit these bugs out by hand. It looks like the problem has to do with
the SoShapeHints node - vertexOrdering and shapeType. I think setting
shapeType to UNKNOWN_SHAPE_TYPE (turning on two-sided lighting) did the trick.
If you want to figure out where to edit things, the Walkthrough Pro
convertor is nice because it adds in the layer names in the right places
and they end up as Label nodes in the Inventor file. So I would recommend
putting all similar colored objects into separate layers or grouping as
many things into layers as possible. It makes adding and editing URLs,
colors, etc. much easier if you have to do it by hand.
Textures of any sort are not converted (which is a pain since
I like to use lots of textures and especially decals, textures with
a transparent index), so I've tried to add them in by converting PICT
textures to raw RGB with convert from ImageMagick and other Mac-based
utilities, but found by trial and error that this RGB was not the same
as the IRIX file format that was expected. I ended up using SGI's
fromgif program to convert the textures into "IRIX" RGB.
Then I added in and wrote the texture nodes by hand, using
a combination of gview and vi, figuring out the coordinates by trial
and error. I never get it to work exactly right, but I come close.
It's kind of like trying to write a program using ones and zeros.
I convert the Inventor files to VRML by opening them in
WebSpace on an Indy and saving them out as .wrl files. I can't seem
to add in my own viewpoints, background color, or URL manually - no
matter where I place the nodes, even if it looks right, WebSpace barfs
at me. Even if I cut and paste from another world, it doesn't work.
What in the world is going on? (pun intended)
Anyway, the short of the story is:
* Good, cheap (WYSIWYG) tools are needed. We're still in the
"silent" era, like when you had to encode MPEGs from
QuickTime using half a dozen different convertors. When is
a Walkthrough Pro to VRML convertor going to appear? How
about a VRML to Descent/Doom WAD/Terminal Velocity convertor? :)
That may at least get some commercial interest...
* In designing worlds, keep the polygon count and texture size
low if you expect people to be moving around in it. WinWebSpace
is much too slow (although a very nice effort). I want a smart
cheap modeller that knows about about concave surfaces and
booleans - I end up having to split objects up to simulate these
in Walkthough Pro. I am still testing DXF export from Ray Dream
Designer and Strata Studio Pro, and have not been particularly
impressed by the DXF files made by them so far. In all of these
programs, I have no way of figuring what my polygon count is.
Modellers and browsers should be smart enough to deal with
space-saving shortcuts, like extrusion information and
generating normals when needed.
Anyway, until the hardware everyone uses catches up, we should all
be taking lessons from Clay Graham's worlds:
http://www.sgi.com/Products/WebFORCE/stock1.wrl
http://www.sgi.com/Products/WebFORCE/clay4.wrl
...the "stock" world was used as the demonstration world during
Way Ting's presentation in Darmstadt. I wonder why? :) Because...
...they are elegant, small, have multiple viewpoints and a background
color, and make you want to look around. This is how it should be. Break
large worlds up and connect them with URLs! You know, that client/server
distributed thing. :)
Look at all the space IndyZONE takes up, ugh. Lots of polygons
just to emulate text and slow. Anyway, when people are able to create worlds
like Clay's easily on PCs, and Macs, say a solution under $1,000 to $1,500
US dollars total, and good tools for other UNIX platforms, then I'll say
that cspace is here for everyone. Today's Web media blitzes and video games
have tricked people into thinking that to look good you need really fancy
multiple megabyte graphics. You don't! What you do need is speed
(responsiveness), and a consistent metaphor for your environment. That's
why Habitat (now WorldsAway) was so engaging and why I think things like
WorldsChat are too (even if they're a bit confining - read "proprietary").
In doing Web work, some clients scream if the size of a page and its
graphics gets any bigger than 50K. You can spend some time in trying to
squeeze out a few bytes from graphics, but on the other end of a slow pipe,
it's worth it. Let's try to keep worlds small (say, <= 100-200k, gzipped)
until the rest of the world catches up.
-- Kevin
-- Kevin Hughes * [email protected] Enterprise Integration Technologies Webmaster (http://www.eit.com/) Hypermedia Industrial Designer * Duty now for the future!