And further improvement can be made by reading in just the first 1024 bytes or so, and calculate the md5 from that. Only if those match, you do a full comparison.