"David L. Thompson" wrote:
> 
> I was wondering if anybody wants to give a recap of how VDE2000 went
> for those of us that couldn't be there.
> 

  Great meeting, excellent facilities and organization! 
The schedule was packed 8:30am-6pm both days but well worth it. 
Presentations fell into four categories:

   low-level libraries (like VTK, 3D-MaterSuite, etc)
  
   scripting-language interfaces (IDL, MatLab, PV-WAVE etc)

   visual interfaces (OpenDX, AVS, IRIS Explorer, etc)

   misc. application interfaces


I thought the SCIRun presentation was very impressive. Looks sort 
of like AVS interface with VTK under the hood. All opensource.
They claim to have a highly optimized, well-thought-out data-flow 
implementation that runs well in massively parallel environment and can link
in serious compute algorithms. Too bad they opted out of the showdown.

Will Schroeder from Kitware gave a great talk on how to run a company
based on open-source principles:

o support contracts
o commercial products built on top VTK
o paid to add functionality and features
  etc.

check out Eric Raymond's "The Cathedral and the Bazzar"
and www.opensource.org. He talked some about VTK. 
They have a very neat automatic nightly test for their
software that gives an html report each morning.
VTK is 250K lines of code (100K actually executable).

VTK users were well represented, but nobody
from Kitware was there after Schroeder left. They failed to produce 
anything for the showdown. It was very clear throughout 
the meeting that VTK has some very powerful algorithms
but is a bear to learn and requires some serious programming.
Doubt if they could have generated apps for the showdown
in a short time. There was some interest in creating VTK modules
for OpenDX. We had a student do this with vtkDecimate some years
ago ... I'll try to locate the code if anyone is interested.

The scripting apps were able to handle most of the showdown 
examples nicely, I was impressed, but it took quite a bit of script
code and more than once, when asked by the audience if they
could add a feature, the answer was "we didn't build it into the script".
Hard to say how long it would take to modify. 
Some nice hardware volume rendering from IDL.

3D MasterSuite delivers nearly 1000 classes for CG development.
Is that something to brag about? VRML/Open Inventor under the hood.
They claim VRML is alive and well

Some very nice interactive methods from Almira. move a line segment
through the data and get a real-time 2D graph of the interpolation
onto the line-segment. Nets were generally very small. Not clear
how well animation and looping are handled though they did a nice
job with the showdown data.

Mathworks (MatLab) was low key.

AVS: several admitted that it has a steep learning curve and too many
modules. AVSExpress seems to be a reworked, streamlined version of AVS.
Kubota claims AVSExpress is too hard for novice users and offers a 
Data Visualizer-like interface called MicroAVS that has preset vis 
methods and no programming interface. AVS users were well represented,
but not nearly as many as I expected.

Greg stole the show at the showdown with the GIS multiresolution 
data set (did Frank create the net?). The restitching of the surface patches
got cheers and applause from the audience.  Also, the simplicity of the
solution for the molecular data sequence and the general-array importer went 
over
very well. Some were unaware that DX had gone open: "is this for real?" and 
"is it going to stay open source?". 

Iris Explorer/NAG combination had some nice things to offer. Their streamlines
use Runge-Kutta integration rather than Euler and so were the only ones to
get the true shape correct in one of the showdown sets (I think). 


That's all I can think of. Perhaps others can add some.

 Richard

Reply via email to