A method of updating computer software includes identifying an orphaned computer file in a second version of computer software, the orphaned computer file having a name that does not correspond to any name of a computer file in a first version of computer software and computing symmetric divergence between the orphaned computer file and each computer file in the first version of computer software. The method also includes choosing a file from the computer files in the first version of computer software that yields a minimum value of symmetric divergence between the file and the orphaned computer file and using the file as a reference file for compressing the orphaned computer file.