MrMark:
I imagine there are all sorts of practical reasons this wouldn't work all that well, but that's a clever insight. I fear you may have ruined my weekend."I learn a bit every time I visit the forum.", indeed.
If you were to make a lab of it, what would the error analysis look like? Detection accuracy vs brightness difference?
Long ago the math would have been fun but now I only guess if you took the square roots of both readings (with linear sensors, LDR are log sensors?) they might become somewhat linear-proportional and then knowing how far apart they are could put a unit on the separation and the source?