Maybe this is a difference in terminology, I just want to clarify, as my interpretation of Bill's explaination initially seems opposite. If "tilting" forward is the same as sloping forward (downward) O.K...Positive caster has the upper balljoint behind the lower, the imaginary line would tilt back at the top.
Theoretically, the camber definition is correct, but the imaginary line can not be considered as 0 degrees. The spindle actually locates the upper ball joint further inward than the lower. The hub at zero would have the imaginary line displaying negative camber. The geometry difference is engineered as steering axis inclination, a completely seperate measurement and normally not a consideration during your common alignment. It is the factory method of minimalizing things like bump steer through a variety of possible chassis ride heights and alignment specifications (usually within the min/max recommendations i.e. Camber: -2* to +2*).
As a rule, a vehicle will pull in the direction of the front wheel displaying the most camber (+) and/or the wheel displaying the least caster (-). + Camber seems to pull harder, while - caster (in reasonable values) kinda drifts. Unorthodox combinations of these characteritics (and others) are what makes roundy round race cars stick like slot cars in the turns. The shorter the track the more you want the car to turn itself left, obviously longer tracks, have longer straightaways and "looser" turns. Likewise, less "lefthand favoritism" is desired of the alignment.
Of course, there is alot more to it, I didn't even address the rearend, but you should get the idea. Street cars usually favor the left (within a degree) to compensate for road crown.
Rowdy