I find it odd that we can have "standard APIs" for the full complexity of
relational data, and probably eventually for object database data, but it
is perceived to be impossible to do the same for the parse tree of XML
data. I mean it is just annotated tree structures: it shouldn't be rocket
science (but neither is it trivial).
No, we don't have such a thing yet, because it is not easy to develop and
nobody is willing to stop and think things through. Over time,
organizations like TechnoTeacher and ISOGEN *are* thinking it through. I
don't claim we've got the problem solved, but our direction is already
much more scalable, generalized and rigorous than what we are seeing in
the DOM realm.
Our approach is, we think, the same as the one taken by the relational
database people: first think of a model that supports the range of
applications that we want to support (including editing applications,
repositories, simple read-only processors) and data types that we want to
support (documents, DTDs, schemas, "link maps", vector and bitmap
graphics,... all media). Having defined the model, we need a way to
customize it for a particular application: a schema, just as they have
schemas in the relational and object database worlds. Our schemas are
property sets (the schema language