* Information Engineering Group Departamento de Inform

1
On Interactions in the RM-ODP Guy Genilloud,
Gonzalo Génova WODPEC2005Workshop on ODP for
Enterprise Computing

• Information Engineering GroupDepartamento
  de Informática
• Universidad Carlos III de Madridguy.genilloud
  _at_ie.inf.uc3m.es, ggenova_at_inf.uc3m.es
  http//www.ie.inf.uc3m.es/

2
Introduction

• The RM-ODP foundations are extremely valuable for
  understanding some issues with modelling
  techniques.
• e.g., Subtyping vs. inheritance
• e.g., Extends, in Use Case modelling
• Unfortunately, it is difficult to convey ODPs
  wisdom to people used to other modelling
  frameworks, e.g., UML.
• fundamental differences in terminology and
  approach
• lack of explanations and references
• One essential difference is about actions and
  communications.
• Many (OO) modelling techniques have only
  localized actions (every action is associated to
  just one entity), and all communications are by
  messages. (e.g., activity diagrams and
  interaction diagrams in UML)
• ODP has a much more general approach it has
  interactions.

3
Interaction is a poorly explained concept

• The RM-ODP Foundations document is a difficult
  read, and it is only partially covered by Part 1
• Some essential concepts are not defined at all
• e.g., specification, model, relation
• The concepts of interaction and environment are
  ill-defined and/or poorly explained.
• interaction is defined on the basis of the
  environment of an object

4
The definition of Environment

• 8.2 Environment (of an object) the part of the
  model which is not part of that object. NOTE -
  In many specification languages, the environment
  can be considered to include at least one object
  which is able to participate without constraint
  in all possible interactions (see 8.3),
  representing the process of observation.
  ODP-part 2
• In appearance, a simple definition BUT
• The concept of model is not defined, and not
  really explained. So, what is the part of the
  model ?
• why not write instead all the other objects in
  the model ?
• possibly with an explanation that not all objects
  are specified
• The note is more puzzling than helpful. Further
  explanations are required, or at least a
  reference

5
The definition of Interaction (1)

• 8.3 Action Something which happens. An
  internal action always takes place without the
  participation of the environment of the object.
  An interaction takes place with the participation
  of the environment of the object. ODP-part 2.
  
• Apparently, the classification is clear BUT
• It depends on the concept of environment, which
  is poorly defined.
• And what to think of Note 4?
• An object may interact with itself, in which
  case it is considered to play at least two roles
  in the interaction, and may be considered, in
  this context, as being a part of its own
  environment. Def. 8.3 Action, Note 4
• Doesnt it contradict the very definition of
  Environment?

6
The definition of Interaction (2)

• Peter Linington pointed out to us that Note 4 is
  metaphorical.
• An object may be considered, in this
  context, as being a part of its own environment.
  Def. 8.3 Action, Note 4
• OK, but then, an object is never a part of its
  environment.
• So, either Note 4 is wrong (but we do want it),
  or
• participation of the environment is NOT the right
  discriminant.
• No matter how we look at it, we have a
  contradiction
• the definition of interaction is invalid.

7
An easier concept Joint Action

• The fact that an object may interact with itself
  complicates explanations immensely.
• We therefore suggest to first define an auxiliary
  concept, which we may call joint action.
• Definition A joint action is an action in
  which two or more objects participate, each
  playing one or more roles in the joint action.

8
Interactions and Roles

• Interaction generalises Joint action.A single
  object may fulfil all the roles of an
  interaction.
• Interactions, like joint actions, have roles
• An object may interact with itself, in which
  case it is considered to play at least two roles
  in the interaction, Def. 8.3 Action, Note 4
• Is the presence of roles the discriminant of
  interactions (vs. internal actions)?
• No.
• The discrimination is not absolute.
• We need to discriminate actions with respect to a
  given object.
• An action may be an interaction of some objects,
  but an internal action of another object.

9
Internal Joint Actions

• Consider JA1, a joint action between two objects,
  O1 and O2
• Being a joint action, JA1 has roles.
• Now consider O3, which results from composing O1
  and O2
• O3 is a composite object O1 and O2 are its
  components.
• Depending on how the composition is made, JA1 is
  either
• an internal action of O3
•  Hide JA1 in Comp(O1, O2) 
• or an interaction of O3
• JA1 is not hidden

10
How do we know that an action is an
interaction?It is a Design Decision

• In his WODPEC04 paper, Peter Linington explains
  that
• interactions are allocated to interfaces by
  design decisions.
• A similar explanation goes for classifying
  actions
• It is a design decision, whether a joint action
  among components is an interaction of their
  composite, or one of its internal actions.
• Importantly, the design decision is also
  dependent on the specification language in use.
• A Hide operator applies to the terms for actions
  in the specification, or even to the types of
  actions
• Actions are hidden in groups, not individually

11
What is an interaction of an object?

• Involvement of the environment represents
  observability. Thus, interactions are observable
  whereas internal actions are not observable,
  because of object encapsulation. Def. 8.3
  Action, Note 5
• Observability might be the discriminant we are
  looking for, but it cannot be explained in terms
  of the involvement of the environment.
• An action may be unobserved by the environment of
  the object (e.g., in CCS, if an object interacts
  with itself, the environment cant)
• But it may nevertheless be considered as being
  observable, because the objects specification
  says that it is (it is a design decision).
• What is necessarily true of an interaction of an
  object?
• It has multiple roles
• It is observable, in the above sense.

12
The ODP Definition of Behaviour

• 8.6 Behaviour (of an object) A collection of
  actions with a set of constraints on when they
  may occur. ODP-part 2.
• There is a problem with this definition
• Obvious, if one remembers that an interaction has
  multiple roles.
• One needs to know which role(s) an object
  performs, otherwise he/she does to know its
  behaviour (what it does).
• It is not the same thing to be the buyer or the
  seller in a sale.
• Peter Linington made an interesting observation
• the constraints might be on the roles, rather
  than on the actions directly.

13
The ODP Definition of Interface

• 8.4 Interface An abstraction of the behaviour of
  an object that consists of a subset of the
  interactions of that object together with a set
  of constraints on when they may occur.Each
  interaction of an object belongs to a unique
  interface. Thus the interfaces of an object form
  a partition of the interactions of that object.
• There are problems in this definition
• For knowing an interface, one needs to know which
  roles are in it(the Comp. Lang. has Server and
  Client Interfaces)
• An ODP interface belongs to a single object, and
  an ODP binding binds two or more interfaces. An
  ODP interaction is associated with two or more
  interfaces.
• An object may interact with itselfThe interfaces
  of an object partition its interaction roles

14
Conclusions

• The RM-ODP Foundations is a difficult read
• Many subtle points in the definitions
• In fact, some definitions are flawed
  (Interaction, Behaviour, Interface)
• Explanations, and references are lacking. We had
  to second-guess the authors original intentions
  when writing this paper, which was very
  difficult.
• Writing the RM-ODP Foundations is much more
  difficult.
• Easy to get something wrong, or to miss
  something.
• We invite ISO experts to provide the missing
  explanations (and definitions), rather than to
  just fix the flawed definitions.
• After all, the new definitions might be defective
  again
• The RM-ODP Foundations should become much more
  accessible, for them to be known to, and used by,
  more and more people.