I agree with Eric on this one... in the example below, there is nothing in the structure to say where the qualifier belongs... If you

read the data

sequentially you might have a pretty good guess, but a specification that relies on sequence rather than the structural relationships and clear (explicit?) definitions of components for sense and meaning, is an bit fragile in my view...

Exactly. Relying on ordering in the document is very shaky (CSV vs. XML). Far better to explicitly define how (and to what) the qualifier should be applied.

This could perhaps be handled by statements about statements.

i.e. a. leaf has lobed margins b. leaf has lobed margins with spines c. b is rare

(I believe this is known as reification in RDF. The relationship between two objects can itself be modelled as an object, and thus properties can be attached to it, and additional statements made about it)

Just musing...

L.

-- Leigh Dodds, Systems Architect | "Pluralitas non est ponenda http://weblogs.userland.com/eclectic | sine necessitate" http://www.xml.com/pub/xmldeviant | -- William of Ockham