1. MOF Meta-Models and UML Profiles
Jacques Robin

2. Outline MDA’s modeling level hierarchy Purposes of meta-models in MDA
Meta-Object Facility (MOF): a standard language for meta-modeling
Modeling with software abstractions beyond UML
UML Profiles
KobrA-2
Multi-Agent Simulation

3. MDA’s Modeling Level Hierarchy
Meta- Meta-Model
Level M3
MOF
+ OCL
Meta-Model
MOF
+ OCL
Level M2
Model
UML
+ OCL
AND
-OR
UML
Profile
AND
-OR
Special Purpose
Modeling Language
Level M1
Running
Application
Level M0
Implementation
Running on Platform

4. Example Meta-Model CMOF Meta-Model of Use-Cases (simplified)
inherits
1 .. *
extends
actor
+name
useCase
+title
system
+name
CMOF Meta-Model
of Use-Cases
(simplified)
participates
0 ..*
case
0 ..1
includes
e-Store
UML Model: Use-Case Diagram
OrderItem
ValidateCart

5. Example Meta-Meta-Model
parameter
+direction
attribute
+name
+multi
CMOF
Meta-Model
of CMOF
(Simplified)
operation
+name
class
+name
associationEnd
+name
+multi
package
+name
association
+name
inherits
1 .. *
extends
CMOF
Meta-Model
of UML
Use-Cases
(Simplified)
actor
+name
useCase
+title
system
+name
participates
0 ..*
case
0 ..1
includes

6. Purposes of Meta-Models in MDA
Define modeling languages
Their abstract syntax
Their formal semantics
Define source and target anchors for model transformations
What about APIs and Libraries?
They are all written in a given language
They are thus best viewed as built-in model elements to reuse than as part of a modeling language
Thus they should be part of platform models, not of language meta-models

7. What is Meta-Modeling? Meta-Modeling vs. Modeling
Similar formalisms but different purposes
One models applications and meta-models languages and formalisms
Meta-Modeling vs. Ontology Engineering
An ontology is a domain knowledge model, not a meta-model
But a domain is an intermediate level of abstraction and generality between application and language
Meta-Modeling Methodologies
Uncharted territory
Should integrate and reuse principles and techniques from:
Application modeling (formalism similarities)
Language design (purpose similarities)
Ontology engineering (process similarities)

8. OMG’s Meta-Object Facility (MOF)
Key idea:
Instead of defining entirely new languages for the M2 and M3 levels
Reuse structural core of mature, well-known, well-tooled M1 level language (UML Infra-Structure)
Advantages of MOF over traditional formalisms such as grammars to define languages:
Abstract instead of concrete syntax (more synthetic)
Visual notation instead of textual notation (clarity)
Graph-based instead of tree-based (abstracts from any reader order)
Entities (classes) have internal structure and behavior (strings do not)
Relations include generalization and undirected associations instead of only order
Specification reuse through inheritance
Additional advantages with OCL:
Allows expressing arbitrary complex constraints among language elements (more expressive)
Allows defining formal semantics without mathematical syntax

9. MOF Meta-Model Package Structure
UML2 Infra-Structure
MOF Meta-Model Package Structure
MOF2

10. Essential MOF (EMOF) Meta-Model
Minimum bootstrap elements for modeling, meta-modeling and MDA tools
Includes only classes, attributes, operations, packages and primitive types from UML2 Infra-Structure
Does not includes associations, which are replaces by references (properties whose values are classes instead of primitive types)
Any CMOF meta-model can be transformed into an EMOF meta-model
EMOF extends UML2 Infra-Structure with elements to represent:
Instances and their reflective relations with classes
Language extension mechanisms

11. EMOF Meta-Model: Elements Reused from UML2 Infra

12. EMOF Meta-Model: Elements Reused from UML2 Infra

13. EMOF Meta-Model: Elements Absent from UML2 Infra
Reflection package:
Views any model element as an instance of a meta-model meta-class
Provide operations that cut across MDA layers to manipulate model and meta-model elements as meta-objects of their meta-classes
Provides meta-meta-model of a generic reflective API to programmatically manipulate models

14. EMOF Meta-Model: Elements Absent from UML2 Infra
Extention package
Any element can be tagged to extend modeling language vocabulary
Identifiers package
OID: class property with feature isID true
Extent: OID value range
useContainment: when true, all contained elements are added to containing element’s extents
Elements(): returns extent members
URIextent: extent where OIDs are URIs instead of properties

15. Example of EMOF Meta-Model
(simplified)
Class
Attribute
Reference
Actor
+name: String
+inherits: Actor
+participates: UseCase
UseCase
+title: String
+extends: UseCase
+includes: UseCase
EMOF Meta-Model
of Use-Cases
(simplified)
System
+name: String
+case: UseCase
e-Store
UML Model: Use-Case Diagram
OrderItem
ValidateCart

16. Part of UML2 Infra-Structure Reused in CMOF but not in EMOF

17. Tailored Modeling with Software Abstractions Beyond UML
Software abstractions beyond UML can be classified as:
Specializing UML abstractions (i.e., UML2 meta-model elements)
Generalizing UML abstractions
Being unrelated to UML abstractions
Approaches:
No MOF, no UML (the Microsoft way)
Create and use domain-specific or even application-specific meta-modeling, modeling and model transformation languages
Pure MOF, no UML:
Create domain-specific or application-specific modeling language that does not reuse any UML meta-model element but is specified as a MOF meta-model
MOF meta-model reuse operators applied to UML packages:
Define MOF meta-model of new modeling language that reuses UML meta-model elements and add new ones
UML Profile approach (the IBM way)
Define specializations of UML abstractions by stereotyping UML meta-model elements
Can be carried out as follows:
Define MOF meta-model of new modeling language
Associate every top-level element in this meta-model to a stereotype of some generalizing UML meta-model elements

18. Tailored Modeling: Domain-Specific Meta-Modeling and Modeling Approach
Advantage:
Perhaps easier to integrate with Visual Studio?
Drawbacks:
For each new domain or application
Need to redefine entire MDA language infra-structure alternative to OMG´s
Need to implement CASE tools for non-standard languages
D1, D2
Meta-Models
in D2
GUI Editor for
D2 Diagrams
Model in Language D1
GUI Editor for
D1 Diagrams
D1 Model
Repository

19. Tailored Modeling: Pure MOF Approach
Does not use UML
Advantages:
No need to artificially relate new abstractions to UML abstractions
Drawbacks:
Need to define entirely new meta-model that does not reuse any elements consolidated by OMG’s long meta-modeling experience
Need to develop entirely new graphical notation and editors
Need to develop all model manipulation services (code generation, test generation, reverse engineering) that UML CASE tools already provide
D Meta-Model
in MOF
UML
Editor
Model in Dedicated
Language D
GUI Editor for
D Diagrams
EMF
generates
Menu Editor for
D Diagrams

20. Tailored Modeling: Meta-Model and MOF Reuse Operators
Advantages:
Reuses consolidated elements from UML2 meta-model
Drawbacks:
Need to extend UML’s graphical editors to draw elements in D – UML2
Need to extend model manipulation services (code generation, test generation, reverse engineering) of UML CASE to modeling elements in D – UML2
Current MOF operators do not cover UML2 concept generalizing reuse, only reuse “as is” or as specialization
UML2
Meta-Model
in MOF
<< reuse >>
D Meta-Model
in MOF
UML
Editor
Model in Dedicated
Language D
GUI Editor for
D Diagrams
EMF
generates
Menu Editor for
D Diagrams

21. Tailored Modeling: UML Profile Approach
UML2 meta-classes sub-categorized by stereotypes
Sub-meta-class x stereotype:
The instances of a sub-meta-class S of meta-class G are direct instances of S, not of G
The instances of a meta-class G stereotyped by S are direct instances of G, for a stereotype has no proper instances
Advantages:
Maximum reuse the entire OMG language infra-structure
Maximum reuse of UML CASE tools (no need to develop any new tool)
Drawbacks:
Risk to “artificially” recast modeling abstraction as specializations of UML ones
Little to reuse for non-object oriented modeling languages
UML2
Meta-Model
in MOF
EMF
meta-class
Package
Class * * *
Profile
Stereotype
ExtensionEnd
Model in Dedicated
UML2 Profile P
Extension
P UML2 Profile
UML
Editor *
icon *
ProfileApplication
Image
Property
Association

22. Example Profile: KobrA-2 Component
Restricts UML2 meta-model to the subset of elements used as artifacts following the KobrA-2 methodology
Goals of upgrading KobrA-1 into KobrA-2:
Leverage latest MDA standards: UML2 (especially UML2 components), OCL2, MOF2, SPEM
Extend process to PSM modeling and implementation
Turn process more model-driven by:
Clearly distinguishing between CIM and PIM
Substitute all tabular natural language artifacts by OCL constraints and expressions
Formalize process in SPEM
Extend process with sub-steps dedicated to GUI development

23. KobrA-2 UML2 Profile K2CIMStructure K2StructureModel K2CIM
K2CIMBehavior
K2PIMStructureSpecification
K2PIMSpecification
K2PIMBehaviorSpecification
K2PIM
K2PIMStructureRealization
K2PIMRealization
K2PIMBehaviorRealization
nestedArtifact
K2JPMSStructureSpecification
Artifact
KobrA2Component
K2JPSMpecification
K2JPSMBehaviorSpecification
K2JavaPSM
K2JPSMStructurelRealization
K2JPSMRealization
K2JPSMBehaviorRealization
K2JavaCode
K2BehaviorSpecificationModel
KobrA-2 UML2 Profile
K2BehaviorRealizationModel

24. KobrA-2 UML2 Profile 1..* K2StructureModel K2ClassDiagram 1..* * *
Component
Class
Association
Dependency
Interface
1..*
Port
Artifact
Manifestation
InstanceSpecification
1..*
K2ObjectDiagram
0..*
ComponentInstance
Object
Link
Connector
1..* * *
1..*
KobrA-2 UML2 Profile
1..*
K2BehaviorSpecificationModel
K2ContractConstraint
Artifact
Manifestation
0..*
K2StateChart
ProtocolStateMachine *
K2BehaviorRealizationModel
K2ActivityDiagram
Activity
1..* * *
Constraint
Artifact
Manifestation
valueSpecification *
K2SequenceDiagram
Interaction
OpaqueExpression *
OpaqueBehavior
ExpressionInOCL

25. MOF Meta-Model of a Simple Multi-Agent Simulations Modeling Language (MASML)
2..*
Environment
Agent
Sensor
Actuator
1..*
Percept
AgentAction
ReasoningComponent
Agent
ReflexAgent
ReflexComponent
ReasoningComponent
Sensor
Actuator
1..*
Agent
AutomataAgent
EnvironmentStateModel
ModelInitializationComponent
PercpetInterpretationComponent
RamificationComponent
ModelBasedBehaviorStrategyComponent
ReasoningComponent
Actuator
Sensor
4..*
1..*
AutomataAgent
GoalBasedAgent
Goal
GoalInitializationComponent
GoalUpdateComponent
GoalBasedBehaviorStrategyComponent
ReasoningComponent
3..*
EnvironmentStateModel
ModelBasedBehaviorStrategyComponent

26. MOF Meta-Model of a Simple Multi-Agent Simulations Modeling Language (MASML)
KBAgent
ReflexAgent
ReflexKBAgent
ReflexKBComponent
ReflexComponent
KBAgent
KBComponent
PersistentKB
ReflexKB
context ReflexKBComponent inv Volat ileKB.isEmpty()
1..*
Agent
ReasoningComponent
KnowledgeBase
KBSentence
1..*
KBAgent
KBComponent
PersistentKB
VolatileKB
1..*
0..*
GoalBasedKBAgent
GoalBasedAgent
KBComponent
KBAgent
GoalKB
EnvironmentStateModelKB
6..*
VolatileKB
Goal
EnvironmentStateModel
4 ..*
3 ..*
AutomataKBAgent
AutomataAgent
KBComponent
KBAgent
EnvironmentStateModelKB
4..*
VolatileKB
EnvironmentStateModel
4 ..*

27. EnvironmentStateModel
UML2 Profile for MAS
MASML Meta-Model
UML2 Meta-Model
MAS
Component
isActive = true
Environment
Agent
ReasoningComponent
Component
Sensor
Port
Actuator
Percept
Signal
AgentAction
EnvironmentStateModel
Model
KnowledgeBase
KBSentence

28. Available UML Profiles
By OMG:
Enterprise Application Integration (application interoperability through standard metadata)
Enterprise Distributed Object Computing (EDOC)
QoS and Fault Tolerance
Schedulability, Performance and Time
Testing
Third parties:
Enterprise Java Beans (by Java Community Process)
Software Services (by IBM, supported by Rational Software Architect UML CASE tool)
Knowledge-Based Systems (University of York)
Data Modeling (by agiledata.org)
Framework Architectures (UML-F)
Requirement Engineering with KAOS
Formal Methods in B (UML-B)
Embedded System Design