A metamodel is a model whose instances are models.1 A GenVoca model of a product line is a tuple whose components are features (0-ary or 1-ary functions). An extension (a.k.a. delta or refinement) of a model is a "meta-feature", which is a tuple of deltas that can modify an existing product line by modifying existing features and adding new features. As a simple example, consider GenVoca model M that contains three features a-c:
Suppose meta-model MM contains three meta-features AAA-CCC, each of which is a tuple with a single non-identity feature:
where 0 is the null feature. Model M is constructed by adding the meta-features of MM, where + is the composition operation (see FOSD).
MM models a product line of product lines (PL**2). That is, different MM expressions correspond to GenVoca models of different product lines..
"Scaling Step-Wise Refinement" (PDF). ftp://ftp.cs.utexas.edu/pub/predator/TSE-AHEAD.pdf ↩