I'm currently discovering Docbook, so I don't know if this makes sense, but it seems to me that in the documentation of Docbook, there is already a sort of hierarchy.

First, you have structural elements ("book", "section"...), meta-information elements ("abstract", "author"...), and content elements ("para", "mediaobject"...). These look like 3 basic classes from which specific elements are derived.

Then, in content elements, you have blocks and inline elements (including raw text), which again looks like 2 sub-classes of elements.

One could thus imagine to construct a hierarchy of elements like in OO programing.

Concerning aggregates, like "formalpara", it was build like an aggregate, but why wouldn't it be possible to derive "formalpara" by just adding an extra compulsory element "title" at the beginning?

To my opinion, such an OO design is possible, and could simplify the writing of allowed children. For instance, instead of listing all possible blocks as possible childs, I think that only precising which class of elements are allowed, like only "inline" elements in a "simplepara" would make things easier.

However, I have no idea whether this is possible using Relax NG directly (possibly another XML Schema should be invented?!).