The difference is based on the relative frequency of queries where a is important vs. where b is important. Either way, what will happen is that the index scan will isolate the first item mentioned, then scan the result set to isolate the second term.
e.g. SELECT * FROM someTable WERE a = someValue and b = someOtherValue Step one isolates the matching "a" values. Step two walks through that resultset and examines each value of b. Since the values of b are already in memory, all it has to do is examine the index keys to find the matches of b. No additional disk read is required. Arthur
