Hello, Thanks for your answer, but maybe I haven't been clear enough in my initial message. Splitting the figurine into 20,000 Shape3D isn't a problem at all in my speed issues, since collision detection works fast in that case (after setting the automatic bounds computation): I first test against bounds, and then, if the result is positive, I test against the geometry (I need that for collision response - find the normals, the vertices, and so on...). I have no speed issues until I want to detach this 20,000 Shape3Ds branch group. It behaves as if the garbage collector or Java3D's equivalent (no idea what Java3D really does when detaching a branch group) wasn't able to handle such amount of data, even if allocating the data (by building the scene graph) is very fast!
So it would be hard for me to give up my fast collision detection system only to be able to detach the branch groups in reasonable time, since I'm sure (or at least I'd like to be convinced...) that there is a solution, something like telling the garbage collector "here, I swear you that you can get rid of those data..." I know, this isn't in Java's philosophy! Anyway I will try and think to use a lesser triangulated figurine at least for the collision detection, but I find it a pity to be obliged to do this only to detach a branch group. It may be faster to close the application and start it all again instead of trying to update the scene graph... Nevertheless thank you for your help! =========================================================================== To unsubscribe, send email to [EMAIL PROTECTED] and include in the body of the message "signoff JAVA3D-INTEREST". For general help, send email to [EMAIL PROTECTED] and include in the body of the message "help".
