140

Designing systems that (may) survive the internet

Think Tank 75 minutes

Investigating ways of expressing interaction in a component-based system

Kirby McInnis

Traditionally object-oriented systems have tended to "hard wire" together objects in hierarchies, through inheritance, and eventually applications through collaborations and associations. This approach to solutions engineering was effective when requirements were "stable" and deployment was to a definable topology. The need to create distributed systems, brought on in large part by the internet, along with the rise in customer focused processing such as mass customization, portals and "markets of one" have resulted in a need to create much looser couplings between objects and components. This looser coupling is often achieved through the use of components and script code, where the components reside on many different nodes within and outside of an enterprise and the script code may be dynamically generated based on user preferences.
When the life span of a given system is measured not in decades but in seconds, how best to capture and express the interactions that are planned in the version 1.0 architecture of a solution?

Michael Sparling (msparlin@castek.com)

Castek

Mike Sparling is chief technology officer for Castek based in Toronto, Canada. As a member of the Castek executive team he is responsible for broad strategic and business planning for the entire group of Castek companies. Mike leads the advanced technology and research group, providing technical leadership for the sizable annual investment Castek makes in research and development. Mike's principle fields of interest are application patterns in distributed component computing and intentional and domain specific notations for component specification. It is expected these initiatives will deliver multi-fold improvements in developer productivity and improvements in the accuracy of implemented application systems within the next five years.

Kirby McInnis (kmcinnis@castek.com)

Castek

Kirby is a Senior Technologist at Castek. He has extensive experience with CASE technology and has worked in all areas of the CASE development life cycle. In his current role, he is working at incorporating object technology principles into Castek's own component-based development and their CBD/e consulting product. His primary area of interest is in developing guidelines for the development of components based on the concepts of business process, business objects, and data objects, the goal of which is to develop business components that reflect the business and not the prevailing technology and user interface. Kirby also enjoys contributing to, and acting as editor for, cbd-hq.com, a web page dedicated to discussing ideas about component-based development.

Topics

Portals, web applications, alternate computing devices and enterprise application integration: What does the world look like that our brave little systems have to cope with.
Collaboration, Interaction, Joint Actions, What else?: How can we model cross component/object behaviours and how best do we apply these techniques?
To factor or not to factor: When do behaviours get factored to an interface and when do they remain in the application code?
How can we express a behaviour so that it may be realized in script? Can we automate the generation of the script based on interaction rules? How could we do that?

Benefits

Participants in this session will have an opportunity to interact with other practitioners and researchers who are focused on the topics of interaction modeling, pattern and framework based generation and application deployment in the age of the internet. This is an area of importance to software vendors and business users as they look to post Y2K resurgence in application development and deployment. Due to the topic, this session will be run as a highly interactive and moderated brainstorming session.

Session: Think Tank 75 minutes

Level: intermediate

Audience: This session is best suited for practitioners and researchers interested in analysis and design notation and the creation of distributed applications created with components. Anyone asking the question "how do we design applications for the internet age that are responsive, customizable and tailored to each individual user" would be a candidate for attending and anyone who thinks they know the answer to that question should attend.

Max 15

Material

Since this is a real world problem we face in planning our applications, we will prepare a case study outlining the problem as we see it. We will also prepare a paper on what we perceive are the generally accepted way of modeling interaction between objects and components and the short comings we see in these techniques. These materials will help set the stage for the discussions we hope to generate.

Delivery

We will facilitate the session, and act as scribes to capture the thoughts of the group. During the session conclusions we will seek to create a poster detailing the major ideas generated by the session and as a follow-up exercise we will produce a transcription of the session provided all participants are willing to let us record the proceedings on magnetic tape.

Format

This session will take the form of a think tank / brainstorming session, with high user participation. We don't know the answer to this very real world problem we face in our business, though we do have a lot of ideas. We seek alignment with peers researching and exploring the same problem. To reach alignment, we need to have a focused and highly participative brainstorming session.

Structure

Level setting: Components and application architectures. 5 minutes.
Established modeling techniques: What are the generally accepted ways of expressing interaction between objects and components. As a group we'll attempt to establish the applicability of these approaches to the problem at hand. 25 minutes.
Resolving the locality of the interaction: When does an interaction become an operation and when does it remain at the "application level"? Are there guidelines for this? 20 minutes.
Additional techniques for refining the "application level" artifacts for use in script or program code generation. 15 minutes.
Final thoughts and conclusions. 10 minutes.

140