When another car hits an object to which your computer (which is watching the car) which by your computers account has busted a spring the car will proceed without a spring. However if this happens and the other computer says 'nope the springs fine' the GUI for damage updates back to a normal state but the car will continue with a broken spring bouncing around because on the other end the car is actually sitting differentlt because the spring isn't broken.
Uhmmm No... Not to me anyway...
Could you please explain that one again, and break it up into events...
Like...
My computer is watching a car in the race.
That car hits a kerb mighty hard.
My computer says that it broke a spring.
The other computer says everything's fine with all the springs.
My computer now shows the car driving around with a broken spring.
The other computer is updating the position of the car without damage (speed and drivability).
My computer shows the car driving around bouncing like crazy because my physics says it's broken, but the other computer says it's not, so it's going between broken and not broken on my computer...
Did THAT make sense?
More importantly, is that what you mean?
IMO it cannot happen, because online sync. sends info about damage(or it shloud send). So if your computer thinks that the susp. is broken and the other doesn't, the difference will disappear when sychronization packets are sent be another PC any recieved by yours. And By the way, HOW could your CPU calculate a different physics than the other? They are using both the same programme code, aren't they? (It might happen it there were some random generator in LFS, but I don't think it is there)
Do you have some experience with it, or it was only and idea what should happen?
Unlikely. I think you'll find all that LFS sends is location packets, and some extra traffic, but certainly not *visual* damage packets - it would create too much latency - and I think we'll find this is the problem, and that the actual car in question may have *looked* damaged, but if you inspected it, its unlikely to be really damaged (i.e. look at the suspension information, etc.)
If there was random generation in LFS, the server would deal it out, so that point is NULL and void.
Racing games are a little more tricky than online FPS games, or anything else, because of the massive amount of things that can alter. Attempting to sync these all with individual packets would be astronomically difficult and very very bandwidth intensive - so I think you'll find that LFS updates car position more often, and the obvious client side prediction, in the hope to avoid this issue.
Unfortunately lag, or other external influences mean that the client side prediction isnt always right. Whilst the server may never see the car damaged, a client may do. However due to sync and snapping the visual damage may remain. As for real damage remaining (i.e. data which affects the simulation), its re-sync'ed and it carries on working, but the car deformations cannot be undone (its too complex to do, unless you "save" a visual deformation state, before dealing damage - so you can roll back).
