by Spike » Thu Feb 12, 2015 9:46 pm
the difference between Q1BSP and Q2BSP collisions is simply this:
Q1BSP is just a load of planes.
Q2BSP has the planes stored in convex brushes.
that is, Q1BSP uses the BSP tree itself for the collision, while Q2BSP merely uses it to find the appropriate brush.
This means Q2BSP can expand the size of the brushes for the collisions (after walking the bsp and following more nodes than strictly required), while Q1BSP cannot and thus can only support point-sized collisions.
Which is why Q1BSP actually contains 3 separate BSP trees (HL contains 4, H2 contains about 6). Each BSP tree is build from pre-expanded brushes, which is why they can't be changed other than by hacking the qbsp and recompiling.
.