Video follow-up

I was really pleased with the turnout for Gregor Hohpe's Developing in a Service-Oriented World. The presentation itself unfortunately (for a business audience) includes a lot of computer science arcana, so in follow-up here's a bit of explanation of some of the more inscrutable points:

• Prolog. Prolog was, for a while in the '80's, the poster-child declarative language. You didn't exactly program prolog; instead you entered "facts" and "inferences", and then used a query to tell the program to solve itself. Prolog's origin (France) and theoretical base (Horn clauses) precluded any chance of widespread adoption. As a development metaphor, prolog really does make sense for us, as a Tranzax system does (like prolog) build facts (elements) and inferences (rules), and runtime execution flow is determined dynamically. Prolog is worth some study if you want to understand (conceptually) where our products are headed.

• MDA: an appealing but (IMHO) wrongheaded formalism that underlies much of visual programming today. Originally championed by the Object Management Group, MDA presupposes that the proper design of a programming model yields (as if from the head of Zeus) the right architecture -- in their words "function follows form". I myself much prefer....

• ADM: The idea here is that we are modeling human systems, which themselves arose organically and stochastically. We capture what actually happens at desktop-level, and the architecture follows. Form follows function. Doesn't that feel better?

• Starbucks Does Not User Two-Phase Commit. Great example - a simple process that everybody understands that is nonetheless complex enough to clog a boxes-and-arrows process modeler. The killer idea here is that many optimized processes are inherently asynchronous, and a process model that enforces synchrony with result in a less-optimized process than the original. Hohpe's writeup in the link is interesting and worth reading.

There's a lot to the Hohpe presentation -- please let me know if there are other concepts that need to be explained further.