Abstract
This short paper is an essay on the drawbacks of model
driven software evolution which apply equally well to
model driven software development. The idea of
automatically generating code changes from a UML
type model is equally enticing as that of automatically
generating whole components from such a model. The
drawback is that there is then nothing to test against,
since there is only one description of the system, the
model. This violates the principles of software
verification and validation, according to which
correctness can only be demonstrated by comparing
two independent descriptions of the same solution. For
this reason, the author proposes another interpretation
of model driven evolution, one in which the
requirements model serves as a basis for propagating
changes to both the code and the test, along two
independent paths. The UML type system design could
then be generated from the code and not vice versa
Users
Please
log in to take part in the discussion (add own reviews or comments).