There are a couple of threads here relating to collision issues at the moment.
Most indie enines have suffered badly in the past with such collision problems and most still do.
We have in FPSC world objects which by and large are composed of relatively few polygons and characters which have many complex polygons. This can result in its simplest interpretation with characters and perhaps other entities moving through solid objects because the world objects have few polygons which are widely spaced apart. Characters which can move partly or wholly through walls or player which can fall out of level. Characters and some other entities have many polygons which can get caught up in the world object polygons and the player can get stuck. Characters can get stuck or may not be able to get through doorways because their box collision area is too large. Characters can fall through floors or embed in floors and sometimes apparently dissapear from levels.
There seems to have been many variations to FPSC models and packs and some inconsistencies. Add to this users too utilise their own or external models in their levels. Characters and other entities can be affected because of their design in a modeller at creation time, the engines model positioning system and FPSC understanding of their central axis vertex, bone positoning and so on. Badly designed, corrupt or models with variations to the .x file description language which FPSC requires can all cause collision and other problems. All in all collision is a complex subject with many factors involved and influencing the result.
The collision system used for characters in FPSC is of the simplest variety. FPSC does not use polygon collision for characters so it is inaccurate.
A quality collision system for FPSC would require a considerable amount of additional math to take place inside the engine at run time. Polygon collision requires a great deal of math. These things are an expensive drain on processor and engine calculation requirement and time. FPSC is not a powerful engine in this area and does not have spare capacity to accommodate the extra math that would be required. It has none to spare. The extra calculations required would incapacitate FPSC and result in even slower fps and more lagg than you have now and stand still in game play.
Almost every factor that affects fps and gameplay speeds affects quality of collision. As fps drops off quality of the collision system deteriorates and collison problems increase quickly as fps drops off. You can see this happen easily in situatons where fps is low and lagg increases and characters behaviours become badly affected. The likelyhood that their collision becomes affected greatly increases and can be seen quite clearly taking place.
It is not only the collision system but a combination of the many other well knowwn factors (AI drain, the high poly counts and errors generated in compile, lack of poly culling) in FPSC which cause an apparent increase in collision problems.
It is difficult to adjudge how much influence these additional factors actually have on quality of collision because they cause instability of behaviours at run time.
Some users are asking for priority of stability for FPSC and certainly in theory at least the more factors that are removed from the sphere of influence the better should be performance in other areas.
The removal of every issue also makes it easier to analise, track down and isolate the root cause of other issues and provide a fix.
Something of a process of elimination which eventually may return greater stability.
Of course their is a limit in any engine dependant upon the quality of the base engine and source code and any external plugins and their source code also. An exetrnal physics or collision engine being typical examples.
Complex stuff Yes.
"There are those who said this day would never come - What have they to say Now?"
