Re: aggregation and states



Date view Thread view Subject view Author view Attachment view

From: Ulf Schuenemann (ulf@cs.mun.ca)
Date: Thu 13 Jun 2002 - 19:16:54 BST


Yes, as Mr Henderson-Sellers said via Andy Evens [thanks for the
references], "a lot" has be written about whole-part (in object-oriented
modeling and otherwise [sidenote: outside OOM, one seems to prefer
"part-whole"]):
what should count as a whole and what as a part and what not, and what
kinds of wholes, of parts, and of whole-part relations there are.
The term "aggregation" is used by a variety of authors for a more or
less precisely described kind of whole-part relation.

But, coming back to Jonathan's original question,

  "how [UML?] aggregations constrain behavioral states?"

[I assume this means UML's aggregation, since this is a UML mailing
list, and not a general object-oriented modeling techniques one]

The component/integral-whole kind of part-whole relationship has been
defined by a functional dependency between whole and parts.  This may
imply constraints on states, but I do not remember any explicit
statements about this.  Behavior, in particular, seems no direct
concern to authors of parts & wholes (OOM and non-OOM). Unfortunately.

From the state modeling side, Harel (father of StateCharts) and Gery
["Executable Object Modeling with Statecharts" Computer 30(7) July
1997, early version in ICSE'96] are only concerned with composite
objects.  They accept (weak) aggregation only for consistency with
UML.  They describe the semantics of UML's aggregation (wrt. state
modeling?) as being ``essentially just that of a special
association.''  Preferable would be a more explicit statement like
"aggregation is irrelevant to state modeling" (but maybe it isnt't?).
[NB also for composition Harel+Gery do not mention constraints on states]


Ulf

Date view Thread view Subject view Author view Attachment view