Your proposed algorithm is a particularly poor solution in terms of correctness if nothing else. For text, DMP performs quite well- patches can typically be applied in arbitrary order with the same result. For complex data formats, you very quickly arrive at invalid (JSON).