21 Jul 13, 06:31PM
OK, lets try summing things up and hopefully provide answers to anyone wondering what the hell all this fuss was about and of course to Boeufmironton, who actuallly asked for help on this matter.
General rule:
Models should always be as low poly as possible and reasonable.
Special case - single plane models:
Using only one single plane (4 vertices/2 triangles), most likely in conjunction with a masked texture, is the method for creating models with only as few polygones as technically neccessary. Therefore it's the number one method for keeping the polycount as low as technically possible.
Unfortunatelly there are several downsides, which have to be considered and may happend to be good reasons for using more polygones instead:
- Shoot through issue no. 1:
The collision of bullets on a plane with a masked texture applied isn't as accurate as on 'fully textured' meshes. (lucas please correct me if I'm wrong) Whether shots pass or not therefore happens to be experienced as something random/unpredictable.
- Shoot through issue no. 2:
Usually the backface of a polygone is not rendered and therefore doesn't stop bullets and doesn't block the view neither. This can be abused by creating walls out of those models, which you can't see nor shoot through from one side, while you can both from the other.
=> Add a second plane to the backface of single plane models, rather than setting the mdlcullface command on the model's config file!
- Volume issue:
A single plane is two-dimensional, in other words it has no z-dimension. It will practically disapear when viewed from certain angles. This may cut off the 'usability' of a model quite strictly.
- Lighting/shadow issue:
Admittedly the least issue to care about, though maybe quite of importance in particular cases. Shadows are calculated on polygone level, not on pixel level. (again, lucas please correct me if I'm wrong) Masked textures won't be taken into account in this regard and the plane will always cast a square shadow, no matter what the actual rendered pixels look like.
On a side note: This would become more concering if shaders were available.
Special case - supposedly unnecessary faces:
Sometimes you may want to leave out faces, which aren't supposed to be seen anyways. For example, the fence pole model used to have no caps on the lower end (version 0.93), since it was supposed to always be attached to the ground and not to float in the air. However, eventually mappers started to make different use of it and those missing caps had to be added afterwards.
Conclusion
The prior thing to care about should always be performance and in this respect the load of (additional) polygones or pixels a model would cause.
At the same time you've to keep an eye on matters of resemblance to the actually modelled object, 'usability' and unfortunatelly even potential threats of exploitation, like explained above.
General rule:
Models should always be as low poly as possible and reasonable.
Special case - single plane models:
Using only one single plane (4 vertices/2 triangles), most likely in conjunction with a masked texture, is the method for creating models with only as few polygones as technically neccessary. Therefore it's the number one method for keeping the polycount as low as technically possible.
Unfortunatelly there are several downsides, which have to be considered and may happend to be good reasons for using more polygones instead:
- Shoot through issue no. 1:
The collision of bullets on a plane with a masked texture applied isn't as accurate as on 'fully textured' meshes. (lucas please correct me if I'm wrong) Whether shots pass or not therefore happens to be experienced as something random/unpredictable.
- Shoot through issue no. 2:
Usually the backface of a polygone is not rendered and therefore doesn't stop bullets and doesn't block the view neither. This can be abused by creating walls out of those models, which you can't see nor shoot through from one side, while you can both from the other.
=> Add a second plane to the backface of single plane models, rather than setting the mdlcullface command on the model's config file!
- Volume issue:
A single plane is two-dimensional, in other words it has no z-dimension. It will practically disapear when viewed from certain angles. This may cut off the 'usability' of a model quite strictly.
- Lighting/shadow issue:
Admittedly the least issue to care about, though maybe quite of importance in particular cases. Shadows are calculated on polygone level, not on pixel level. (again, lucas please correct me if I'm wrong) Masked textures won't be taken into account in this regard and the plane will always cast a square shadow, no matter what the actual rendered pixels look like.
On a side note: This would become more concering if shaders were available.
Special case - supposedly unnecessary faces:
Sometimes you may want to leave out faces, which aren't supposed to be seen anyways. For example, the fence pole model used to have no caps on the lower end (version 0.93), since it was supposed to always be attached to the ground and not to float in the air. However, eventually mappers started to make different use of it and those missing caps had to be added afterwards.
Conclusion
The prior thing to care about should always be performance and in this respect the load of (additional) polygones or pixels a model would cause.
At the same time you've to keep an eye on matters of resemblance to the actually modelled object, 'usability' and unfortunatelly even potential threats of exploitation, like explained above.