Yes, Dr. Mu's concept is exact, but it need to solve collisions, too - generally, problems related with real dimensions of pieces.
Ad intermediate points - imho, my concept of occupied squares is better (but not exhausting collision problems), of course, when it is possible to split every piece into set of square-unit.
Comment on Re^4: Closed geometry: a train track problem
We're pretty much saying the same thing, but I think my points-along solution will serve better when we get to bridge clearances and things like that. For instance, let's say we have a bridge that crosses at a non-right angle. With my version, it's very easy to get a pretty accurate location and angle of the crossing so that we can ensure that the bridge abutments are far enough away from the center of the bottom track to clear. This requires that we know things like the length of the bridge and the width of the bridge as well, but it seems like it would be extremely difficult to figure any of this out without knowing the location and angle of the crossing.
With mine, the allowable distance is the same in every direction, set by one constant. With yours, it's different for side-by-side sqaures from corner-to-corner ones. Since calculating sqrt(delx**2+dely**2) < limit is what computers are supposed to be good at, I rest my case. :D
Don Wilde
"There's more than one level to any answer."