Issues in Moving to a Component-Based Industry
Moderator: Bertrand Meyer, Interactive Software Engineering, USA

This panel will address issues involved in the software world's transformation from its traditional custom-development scheme into a component-based industry. It will discuss technical as well as organizational, economic and political aspects.

New Dimensions in OA&D: UML/OML and Beyond
Moderator: Donald Firesmith, Storage Technology Corp., USA

Although the OMG has officially adopted UML as the industry-standard notation for object modeling, it also has some significant critics. The OPEN Consortium has developed OML as an enhancement/extension answering some of the criticisms that have been raised against UML. This panel will discuss the impact on the object community of the standardization of notation, metamodel, and method that has been taking place the last few years. It will also discuss some of the differences between standard UML and OML, and prognosticate on where the industry is likely to go over the next few years.

OOLs in 1998: Fin de Siècle or Renaissance
Moderator: Richard Riehle, AdaWorks, USA

Over the past 30+ years, Object Orientation has moved from a research topic to a mainstream technology, and OO languages have become the subject of discussion and debate even among lay users and in the popular press. As we are on the threshold of a new century, indeed a new millennium, it is worthwhile to look at what has been accomplished and what remains to be achieved. What are the promises of Object-Oriented Programming? Has OOP been oversold? Does it really make a difference which OO language is used? Are there emerging technological trends for which OOP is a good match? In short, how suitable is OOP (and specific OOLs) for addressing the software development needs of the 21st century?

Proponents of several leading OO Languages (in alphabetic order: Ada 95, C++, Eiffel, Java, and Smalltalk) will address these issues and probably hurl friendly insults at their fellow panelists' languages.

Using Production Rules in Object-Oriented Systems
Moderator: Mark Koenig, Butte College, USA

One of the major stumbling blocks in software development has always been the semantic gap between the "users" view of the system being modeled and the programmers view of the functioning software. Nowhere is this more true than in the manner in which we codify and maintain business rules in our applications. Efforts to close the gap have taken many paths. The focus in this panel is the pragmatic integration of production rules within an Eiffel development framework.

As a software developer and trainer, my interest is in separating the "business rules" written in a very intuitive syntax from the underlying implementation code necessary to execute the system on available hardware. It would be extremely useful if an end user could maintain the business rules for which he or she has authority. This would require a maintenance interface and an intuitive rules syntax. For example, Butte College's registration process would take a form like:

Student_enrollment

student.registered and student.returning OR
Student.admitted and student.new and student.orientation_complete
Instructor.override

Student.enroll
Student.confirm

where an entity like student.registered could be of arbitrary complexity and have all the advantages of the object language.

After we create a clean separation of the "users view" and the design, subsequent advantages for the integration include:

• Integration of the rules and objects such that there isn't a need to map a rules space and its slots with the instance objects of the container system
• Symmetrical rules allowing forward and backward chaining with one knowledge base.
• The ability to modify the production rules without a compile of the container system. Even better, would be the ability to change the production rules in a running system.

In order to close the gap between the "definers" of the system and its "constructors" the object paradigm has used the concepts of abstraction and information hiding. A primary question is then: Can production rules abstract the problem domain in an intuitive form for the domain experts and spare them from the design details? A secondary but connected issue is: Can this "business rules" view of the system be subsequently modified by the domain experts at this abstract level while retaining the integrity of the object system.

Research in this area has been rich, with François Pachet's NeOpus extension to Smalltalk in which the production rules have been implemented completely inside the Smalltalk paradigm. There are obvious advantages to the rules not being a separate product and vendor with it's own methodology.

Other significant research in the Eiffel language include a second level of assertions that would separate the "business rules" from underlying implementation logic proposed by Sita Ramakrishnan in "Two layers of constraints for an extended object model in Eiffel", (Proceedings TOOLS USA 94), and Group S in Belgium using a hybrid system to write complex payroll systems.

It would be of great advantage to Butte College to have a practical implementation of a production rules system that fit well with our object development projects. The type of development undertaken at our site is, I think, characteristic of a large segment of the general business community in its pursuit of manageable business systems.

Refer to the Conference Schedule for more scheduling information