Metainformation about a document element or domain object. In previous iterations of my workflow these were attributes rather than child elements. I generally favor attributes for metadata like this, but an element can have only one attribute per namespace:name. There are some shenanigans to get around this, such as having a ‘aka’ (Also Known As, name aliases of the element) attribute with multiple values separated by semicolons, but these need to be unpacked any time I need to use them.
I moved away from this a while ago because of the cardinality complications involved in some places, such as subdomains associated with multiple domains, or archetypes that can apply to more than one class. Having a blended model with some metadata stored in attributes and some in child elements seemed like a recipe for pain, so moved entirely to child elements.
… and now I wonder if they should in fact be child elements, or if I should treat them much as crossreference tables used to implement many:many relationships in relational databases…