Object behaviour description and understanding
Object behaviour description and understanding
Our recent work has focused on meta-level architectures and
environments for describing and understanding object behaviours. The
general idea is to create an "object engine" on which many different,
wildly diverse object models can be implemented, composed and
otherwise interact in meaningful/useful ways. This work can be
subdivided into several subprojects:
- CodA, an object meta-level architecture
which features decomposition by logical behaviour as its
primary design guideline. CodA has been implemented in Smalltalk and
is the basis for all our object model designs.
- Tj, a distributed object system
implemented in Smalltalk and on top of CodA. It features: transparent
object location, object replication and migration, computation
migration, flexible type-independent object marshaling and execution
in large-scale (~256-1024 processor) distributed memory environments.
- PortedObjects, a dataflow object model in
which objects form arbitrarily complex computation nodes in dataflow
graph. This model is useful in data analysis applications, visual
programming, and system configuration. Its implementation (again in
Smalltalk and on top of CodA) shows some interesting uses of the CodA
meta-component architecture.
- Vibes, a data analysis and visualization
toolkit based on PortedObjects, allows the composition of analysis
functions into an analysis graph in much the same way as IRIS Explorer,
AVS and Pablo. One key difference being that the type of information
to be analyzed is not limited to 'data' (i.e., numbers).
Several documents relating to many of
these projects are available on-line.
|m 95.02.01