Impossible? Are you sure? I think more things can be done with bricks in other engines.
http://www.youtube.com/watch?v=qZHPT6i4QfA
Did you link to the wrong video? Cause all I see there is maybe 200 physics objects that glitch out and disappear when too many of them are disturbed at the same time.
The problem here is that you guys don't understand what a game engine provides for you. There is no game engine that will be super fast at rendering 100,000+ discrete objects out of the box. You cannot simply fire up source or crysis, stack up a million boxes and go to town (those crysis videos that are all over youtube are pre-rendered). If you want to do something unusual like that, you're going to have to write your own renderer and you're going to have to cut corners on the simulation. There is no general purpose "have a million of everything" engine.
Cutting edge engines like crysis, source, unreal 3, and id tech 5 cost upwards of
a million dollars (still way cheaper than developing your own engine) and even if you have the money (I don't) they won't license to everyone because they only want to associate and give tech support to AAA titles that have a reasonable chance of success. Second tier engines like Unreal 2 or doom III cost in the $250,000 range (I don't have that much money either).
I need an engine that is priced and designed to be used by a single person or small team. Here are the possiblities: Torque, TGEA, Ogre, Blitz, flash/shockwave, crystal space, C4, unity, Sauerbraten/Cube 2, Quake III GPL, DIY. (Sauerbraten, TGEA, and unity did not exist at the time I started Blockland)
Of those, Sauerbraten is the only one that could possibly provide a graphical advantage out of the box because of it's unique rendering paradigm. BUT. There are glitches in the rendering system that look quite difficult to solve and may be systemic. There is no scripting system. The networking sucks and would have to be replaced completely with something like TNL. I've talked about it with kompressor and we concluded that it would be easier to port the sauerbraten rendering over to torque.
I picked torque because it's the only one with a good scripting integration and robust networking that works well with low bandwidth. It also runs well on low end hardware and can be ported to the mac, appealing to the widest audience possible.
The most important thing about making a game is having it work. Jumping around from one engine to another chasing after the latest features is a sure fire way to fail (are you reading this George Broussard?).
