That's pretty much exactly what I thought happened. The collision detection on that cloth is a rough approximation of the actual cloth. Now that's a pretty good approximation, usually in games like GT the apporixmation is a bit more gross. That is an assumption of course, but with what's going on in GT5, at the rate it has to happen, it seems like collision detection even on on the level of the tablecloth you linked is beyond reason.
This is just a quick illustratrion of what I am talking about:
This is how collision detection is handled in an FPS. A colision of the bullet with a box is registered as a hit. It's not terribly accurate but it gets the job done with a lot less work than actually detecting every single poly in the visible model.
Now imagine draging that glob of hitboxes over a surface and having leave behind damage. Would it leave behind nice smooth human body looking damage? No it would look all weird and hard edged. Now imagine that other surface was a jumble of hitboxes, the result would not look right at all. With cars it's a little more simple and I assume the boxes used are less gross in accuracy, however the point remains, look at the texture damage on the cars, it's so accurate looking for damage to an nice smooth complex surface being dragged over another smooth complex surface, it just doesn't make sense it's dynamically generated by dragging to course collision detection models over each other. That's why it seems it's a premade texture, the event that is requriedin an accuracy high enough to yield a good result doesn't actually happen.
The reason I brough that up is if you really have two cars that scratch each other as you were saying (ie the texture is scratched off or a texture that looks like missing paint is painted on) and you use hitboxes (or whatever you want to call that technique) you don't ever have a real collision of the visible parts of the car so you can't calculate what a real scratch would look like.
Look at the damage to the cars, the scratches. The paint damge wraps elegantly around complex parts of the car, into crevaces etc. If you use hitboxes (or whatever you want to call it) you only know the direction and approximate area of contact, you cannot from that draw a detailed scratch or damage because all you really have is two rectangles hitting and the bodies complex shapes never even make contact with each other.
This is how it's usually done with hitboxes:
Collision is detected.
General area is calculated (front, back, which side etc) and from that a damaged texture is swapped out for a pristine texture on that part of the model.
This is why any contact on the front left fender say results in the same exact damage texture in different situations.
GTAIV has has a kind of dynamic damage however it's not very accurate. For instance when you hit another car, the correct panel deforms in a generally correct angle and level for the impact, but it's clearly not an imprint of your car on that car, it's just generalized.
For instance you hit the back right of a car, the entire back right quarter is messed up, but from the center to the far left is pristine. Also even if you hit it coming at an angle, often you see the back smashed in as if it had been hit straight on from behind.
As for seeing it... we can't be absolutely sure, but as someone who follows the happenings as they are reported I am kind of surprised you don't. As it was happening it all seemed to be very much in line and even Jordan came to the same conclusion...
While I am well aware there are a lot of drawbacks in development... we are looking at a very long dev time already, it just seems like 5 years should already include a lot of leway for drawbacks. Everything from internal marketing to KY's quotes seems to point to an xmas 09 release with more or less what we saw at GC and then a sudden pushback to include the things we are now hearing about and seeing (damage to all cars, weather, night races etc). I have been on some projects where very similar things happened, we promised go live on Jan 1 and on Dec 1 we realized we were either going to have to put out a less than impressive result on Jan 1 or push back... we could get a decent product by Feb but if we took until June we could really focus on a lot of good stuff.
Choice: Delay a little to give what was originally promised which makes us just look late - or - delay a decent amount but add a noteable amount more to really justify a delay.
