[osg-users] OpenGL 3.1 at GDC

Jan Ciger jan.ciger at gmail.com
Mon Mar 30 02:25:50 PDT 2009


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Robert Osfield wrote:
>  Another thing that SGI used to do
> was promote the use of scene graphs to solve this problem but moving
> software vendors up the food chain so they could take advantage of all
> the know how and avoid the pitfalls.
> 
> This is where the OSG would come in these days :-)

Honestly, Robert, I am not that optimistic. More often than not the
reason why the crappy code is kept is the not invented here syndrome
(often tangled with obscure legal reasons and worries, especially about
OSS licensing) and  "it works, why to replace it" (implying costs)
mentality. Finally, when you have an application where the 3D is only a
bolted on add-on seen as a mainly marketing driven sales device and not
core functionality you get what you get. E.g. AutoCAD - most of the work
is actually done in 2D, essentially copying the work processes of good
old ruler and compass drafting, so improving the 3D is a very low
priority for them.

Finally, it doesn't help that good engineers who actually have a clue
about these things are scarce and expensive. There are plenty of
"game-oriented" curricula that teach students how to make pretty models
in Maya or Max or how to bang up a bit of code to draw their model, but
that's it. More in-depth study is lacking, because it is not attractive,
neither for the students (math, theoretical computer science needed) nor
for the universities. If you are paid for the number of students
graduating on time, you do not have much interest in teaching hard
topics where it is normal that not everyone will graduate. The result is
a severe curtailing of the teaching (my own experience as a university
teacher), letting you to really show only glBegin()/glEnd() as an easy
way to get the students started while explaining other topics
(transforms or lighting, for example) but no time to move to more
advanced things - vertex arrays, VBOs, shaders, etc. You can only warn
the students that the approach shown is very inefficient, but if
something is not shown in class (due to lack of time or whatever
reason), most of them will never go read up on it.

Unfortunately, this goes also for math and computer science teaching - I
had a hard discussion with a comp-sci master student recently. The
fellow was not happy to have to do a miniproject for exam. After
questioning why I discovered that the guy doesn't actually know how to
program (master comp-sci student!!!) and for him "computer science is
not about programming." For math it is regular occurrence for me to find
students who are incapable of working with fractions, going even so far
as to say that my result cannot be correct, because they get always
decimals on their calculators! Also, I have encountered a computer
science master student during computer vision exam (different one than
above) who couldn't multiply matrices - saying that he never needed it
for anything. And these are regular occurrences at a technical
university, not once in a year flukes.

Are you still surprised at the quality of the code in commercial
applications? I am not.

Regards,

Jan


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mandriva - http://enigmail.mozdev.org

iD8DBQFJ0JAdn11XseNj94gRAtEoAJ9FKTQ6oPCB4tLsL+39i1gckQiJ4wCfUlai
TYtGHr2EyG1k3vRVAOoK2nA=
=6/JA
-----END PGP SIGNATURE-----



More information about the osg-users mailing list