Flexible as it is, XML is an ephemeral format. An XML document holds a snapshot of data, unlike a repository such as CVS/RCS which can store a history of such snapshots, a versioned archive. Such an archive format has many uses, providing an auditable record of revisions, a switchable set of configuration options or a document allowing "track changes" for editorial control.

Ideally such an archive should itself be expressed in XML, allowing the archive to be processed by the same tools as the data it contains. A CVS revision history cannot be transformed by XSLT, cannot be referenced with XPath or pointed to with XPointer.

There is currently no generic solution to this problem.

What are the challenges? The multi-version XML file should of course record just the changes and share unchanged data between versions. Handling attributes presents a particular problem because each may be present or absent in a particular version and could have different values for different versions. Ideally, when dealing with documents changes should be recorded at the word level. It must be easy to extract a version, add a new version into the archive, delete a version from the archive or clone a version within the archive. An overriding requirement is to keep to the spirit of XML, i.e. that the archive is not only easily processed by computer but is also inherently understandable when viewed in a text editor.

This paper proposes a generic XML format to handle multiple versions of an XML document, using the same basic structure as the files which are being archived. The paper discusses the technical issues and challenges, and alternative approaches. Benefits of the proposed format include the ability to extract any version with a simple XSLT stylesheet, and similarly to use XSLT for deletion of a specific version and cloning a version for minor modifications. The most interesting challenge is how to automate the addition of an arbitrary new version into the multi-version file.

The proposal is built on many years of research and commercial experience in identifying and handling changes to XML data. The proposed format will enable development of a new class of applications, which are briefly discussed.