A method and apparatus for representing three or more related markup language source files in a single delta file is provided. The markup language source files are represented in such a way that the delta file is easily processed to produce one or more merged combinations of the source files. The method and apparatus ensure that any one of the original files can be extracted from the delta file. Moreover, when one or more documents are deleted from the delta file, leaving two or more documents within the modified delta file, the remaining file is a valid delta file for any remaining documents. The method and apparatus are suited for use, but not exclusively, with XML documents/files.