XML Digital Signatures
Dr. John Boyer has an interesting article about XML-Signatures.
It's a highlevel kind of introduction but also has some great insight into some of the pitfalls when making XML-Signatures legally binding.

But making digital records with the legal enforceability of paper can be challenging because software developers have been trained to separate logic, data, and context (presentation and questions). Consequently, often what is digitally signed is only the data, or answers portion, of a contract or form.

Later, under regulatory or court scrutiny, it may not be clear exactly what the signature authorized. Did it authorize the purchase of "500" widgets or "500" gadgets? Did "yes" mean that the nuclear power plant inspection was trouble-free, or that further inspections were required? What did the person see and intend when applying his digital signature? It cannot be proven because the context (questions) are not bound with the answers and protected from tampering by the digital signature.

Good point. Ian Grigg has made this point for years and has restisted xml formats for this exact reason. He's always been a proponent of having the signed "contracts" have both human and computer readable content. So it can be upheld in court.
For examples see the contracts at the webfunds site.

I have been designing NeuDist to support XML based contracts. The volume of many of these contracts will be such that you dont really want vast amounts of human readable explanations in it. Being XML you certainly could have it as comments and use the XML Canonicalization algorithm with comments, which I'm doing anyway.

My approach is to have a set of standard certified XSL tranforms for making human readable forms of them. These could be available in different languages as well. It might be a point actually now I'm thinking of it to have that transform be required before the signature. So the version in transport doesnt have the human readable language, but the version that is used to generate and verify the sig does. hmm.