Seperating collision model and displayed model

Instead of having to enforce limits on the size and number of prims that can be linked together and set physical to ease up on calculations, why not seperate instead the collision model from the displayed model ? In many cases a lot of prims used in the models are there only for looks or even meant to be phantom. And in most cases a number of prims can be approximated into one.

So I suggest that SL allows builders to make the collision model for physical objects seperate from the actual 3D model displayed, just like Unreal 2003/2004 does. It would save a lot on server load and allow to build beyond the limit of 30 prims for physical objects.

A similar thing can be accomplished for vehicles now by producing a physical model which behaves as the vehicle would and which provides substance, and then having attachments which provide the detailed model. Unfortunately, this method is a kludge at best.

I highly support this proposal. I think, at the same time, that the way which linksets are handled should be revised. If an object is phantom and is linked to a non-phantom set of objects, I think that the phantomness of this object should persist even when linked. By making this change, it would then be possible to define only the important linkmembers as physical, and only these prims would count against the physical limit. I think this simple solution would be the best way to accomplish what you've described. Actually having two separate models would probably be cumbersome to implement.

When this has been suggested in the past, the Linden response has been that it would require a fairly aggressive redesign of the way linking works... which is good, but that they'd like to finish implementing Havok 2, so as to not have to do it twice. So... don't bet on being able to link phantom and solid prims together anytime soon.

Heh, agressive redesign indeed. Linking is so freaking crappy!
It's not hierarchical, it causes lots of nasty prim drift, linked sets can only be scaled uniformly... the distance limits are RIDICULOUS and 255 prims are NOT enough.
LL needs to realize that building is building, scripting is scripting, physics are physics.
We should have a purely graphical object grouping function completely separate from object-to-object communication.
We should have some sort of group communication system for scripts completely separate from the building interface considerations.
We should be able to define a collision mask completely different from the actual look of the object. A simple bounding box would be enough in most cases, or a collection of bounding boxes around more complex objects.

LL i will kill the person of our choice if you make this feature part of sl. IM me with details then stand back. also all the stuff Eggy said but he has to kill his own people. i'm not going to do the work for him.