Model-Driven Development: A Metamodeling Foundation

In this article we have defined a concrete set of
requirements that an ideal MDD infrastructure should
support, and have argued that the explicit distinction
of two orthogonal forms of metamodeling—linguistic
and ontological—is the key to fixing some of the
problems in the first generation MDD infrastructure
and to scaling it up to satisfy all of the identified
requirements.

The forthcoming revision of the OMG’s MDD
infrastructure in the UML2.0/MOF2.0 standards
represents a significant step forward in this regard in
that for the first time it accommodates two distinct
forms of instantiation. However, two significant
problems remain.

First, although the distinction is present, it is not
made explicit enough. For instance, while linguistic
metalevel boundaries are recognized, ontological
boundaries do not exist. Moreover, when coupled
with the fact that stereotypes are the strongly
preferred mechanism for metamodeling, there is a
strong suggestion linguistic metamodeling is the only
meaningful form of metamodeling.

Second, in the current profile mechanism there is still
a major bias in favor of defining predefined concepts
at the meta-level (i.e. as part of the modeling
language, given the current infrastructure
preoccupation with linguistic metalevels) rather than
as regular user types at the M1 level.