Is it legal to add additional custom information somewhere in an XML prolog? For example, in my case, I would like to add an indicator of which serialiser version was used to create the XML, so that clients receiving the XML could automatically select the correct corresponding de-serialiser.

I could add the information as an attribute of the root tag, but I thought it would be cleaner to add the information in the prolog, like the standard XML version:

<?xml version="1.0"?>

something like:

<?serialiser version="1.0"?>

or is the prolog reserved purely for those things specified by W3C?