A derived model can contain nodes of derived and/or original classifiers

[enikao]

Derived nodes are regular nodes (#254), so they can have properties, references, annotations, and containments.
Which kind of nodes can a derived node contain (or have as annotations)?

We assume #243 does not limit the parent of a derived node.
We assume #181 identifies derived node classifiers somehow.

Option A: No containments

A derived node cannot contain any other node.

Pro:

• Simple

Con:

• Disallows many use cases. Example: a type derivation describes an inferred abstract data type. We want to describe all of the ADT's components, including their derived constraints.

Option B: Only nodes of derived classifiers

A derived node can only contain other derived nodes

Pro:

• Allows use cases like the ADT example mentioned above
• Clean separation between derived and original nodes

Con:

• Might lead to duplication of concepts. Example: Our language allows either explicit or implicit, derived descriptions. A description is composed out of Words. Word might have specializations like ReferenceWord or AnchorWord. For each specialization, we'd need to define a derived classifier and an original classifier.

Option C: Both nodes of derived and/or original classifiers

We can mix derived and original classifiers freely in a derived model.

Pro:

• Avoids concept duplication

Con

• ?

[joswarmer]

If we don't use IDerived, but make derivation a first class citizen we don't have derived classifiers anymore.