Control-theory and models at runtime20104653855
Document Sample


Coherent Support for
Models at Run-Time
through Orthogonal
Classification
Colin Atkinson and Matthias Gutheil
University of Mannheim, Germany
Presented by Olivier Barais and Franck Fleurey
Talk outline
► Introduction to OCA
► Basic principles of OCA
► Multi-level model example
► Open issues
Models@runtime’07 2
Introduction
► With UML, the typing relationship is generally a linguistic
type relationship
It is not so clear in MDE
► This paper proposes to distinguish between two kinds of
“instance of” relationships
A linguistic “instance of” relationship
A logical or ontological “instance of” relationship
Models@runtime’07 3
Basic principles of OCA
► Linguistic level and
logical level are
orthogonal
► Limited to three
linguistic levels
?
► Issue: Why do you
need PLM?
► Could we have
multiple “instance
of” relationship at
the same level?
Models@runtime’07 4
Multi-level model example
Clabject
Fields
?
Models@runtime’07 5
Issues on inheritance and
classifiction of clabjects
Simple inheritance vs. Multiple inheritance
Can a clabject have a « linguistic instance of »
relationship with several clabjects
Simple classification vs. Multiple classification
Can a clabject have a « logical instance of » relationship
with several clabjects
Static inheritance vs. Dynamic inherithance
Can a clabject change its « linguistic instance of »
relationship
Static classification vs. Dynamic classification
Can a clabject change its « logical instance of »
relationship
Models@runtime’07 6
Open Issues
► Benefits of using OCA for models@Runtime
Separate clearly the two « instance of » relationships
Useful @ runtime if you can change dynamically the
logical “instance of” clabjects
What about dynamic typing?
Nevertheless, introduce the same fields at different
levels?
► Do we really need PLM model ?
► Perhaps, could we design a typical models @
runtime app with OCA
Models@runtime’07 7
Get documents about "