The demo You mention, Thomas, only shows bare terrain, no plants, no monsters, no players etc. (but shadows look very good). AFAIK terrain engine in CS works in very similar way to that presented in your example.
i am aware of the fact that it shows simple terrain. but the fact that it shows 160 square km@2.5m resolution is almost unbeliveable, and it sitill has high fps on a average pc.
bdw.. this ranger mk2 thing uses a modified way of SOAR means it re-calculates the entire terrain for EVERY frame. since CS-engine shows popping and few other small things i would guess it to be a ROAM-like which uses single terrain-patches with differnet detail-levels and frustum-like culling.
//edit
i just read in the CS-engine docu that it uses a 4-level static LOD together with splitting the terrain in blocks, each block with independend LOD/texture etc) the visibility calculation is quad-tree based
-so it is somewhat ROAM-like
... come to think about it.-collision detection on such a terrain could be quite a challange.. using the 8*10^9 terrain= bad idea. this cries for somehting else .. well dosent belong here.. sry about that.
//edit end
if you think about... if the terrain showed in the demo i mentioned would be UNoptimised you would have 8.192.000.000 triangles (no this is not a calculation error). i know my graca works still pretty decent with several 10 thausends (guess 40-60k are still ok for realtime) but not 8*10^9. reducing those a little further and you can add lots of trees and houses and players.
Relying that much on specific OpenGL extensions, would increase the hardware requirements further
doing this with vertex shader is something nice but even if you dont have it you could use a software fallback.
using shaders would allow those massive terrains without wastin any cpu power you could use for other calculations.
lightning and shadows.. well per pixel light.. i think you'll have a hard job finding a pc these days without support for this (even most office pc's could do that)
well i dont like to strech this thread. ... crystal space forum dosn't seem to be very active.
i still like to say that this algorithm is ways better and more efficient than the terrain-puzzle blizzard used for WoW or the current CS-terrain optimisation (although already ways those are better than no optimisation).
a mix between puzling and this method to archieve best possible performance would beat anything ever seen.
if anyone is further intrested i like him to join the diskussion on the crystalspace board.
this here is a wishlist and this thread is drifting to a technical discussion which has nothing to do with wishes at all,
best greetings
thommy
ps: waching the long video is realls worth

ps:ps: sry 4 the long post.. bad habbit of mine..