Peter Holditch

Over the past 16 months - has it really been that long?! - I have attempted to climb the peaks of how to design applications that use transactions, and dived into the depths of the earth, looking at obscure knowledge such as how clients can demarcate transactions, and grubby details of interfacing WebLogic's transaction manager with external resources like MQ Series and, much to my surprise, you have come along on this odyssey with me. Who'd have thought that there were so many people interested in these minutiae?

Well, to be honest, I would for one. Being an engineer by training, I like nothing more than tinkering with these kinds of mechanisms. It's interesting, challenging stuff. Some people play chess. I don't know the rules so I need an alternative pointless endeavor! But that is only half of engineering. The world wouldn't contain any engineers at all if they never produced anything - if we all spent our lives tinkering in our sheds, instead of building software systems, bridges, roads, cars, and so on, who would care? We'd be begging on the street corners, transaction manager in one hand, with a mangy dog holding a JDK in its mouth, and no more than an old copy of Kernighan and Ritchie to sleep under at night. That's not the case. We do build things too.

The things we build come in two categories. On the one hand, there is the tedious, repetitive "yet another" forms application that gathers some information from a user, updates some databases, and repeats until bored (or accidentally powered down by the cleaner). On the other hand, sometimes our employers need very technical stuff done, because unless their business is run on some pretty revolutionary systems, with mind-numbing response times and so on, they won't be able to compete and we'll all be out of a job, fighting over the K&R, and who gets first bite of the dog. For the second category of engineering problem, burning the candle at both ends reading transaction articles is time well spent - with a deep and thorough understanding of the subsystems that make up the underlying infrastructure (and a healthy dash of experience and some luck) you can build something truly amazing - the Sydney Opera House of applications.

Who Gets First Bite of the Dog?
However, the Sydney Opera House wouldn't make a great affordable housing scheme. Here builders with more modest aspirations can happily erect a housing complex from prefabricated parts just like they did last week, and will again next week, and make a family or several very happy. What is needed here isn't the pinnacle of creative expression, it's a good job well constructed from components and templates with the minimum of cost and fuss.

So what on earth is he rambling about now, I hear you sigh, has he finally lost it?

Well, not quite. To return from the land of metaphor for a moment, sensible use of transactions will be necessary whether you're building an order entry forms app, or a world-beating innovation. It's just that in the former case, there are many widely applicable and well-understood design patterns for their use (you know, the stuff I wrote about before) that can be applied in the knowledge that they will "just work" because the problem is so similar to so many others. Under these conditions, it should be possible to forget almost entirely about transactions and get on with implementing business logic. After all, how many construction engineers worry on a day-to-day basis about the laws of physics? Transactions really should be fundamental laws-type stuff...

Now, imagine a world where you can graphically describe your business process, add the few lines of code that actually do your processing at the right points, and press "run." Imagine also that you can design (or even generate) the forms front end at the same time. Sounds like second-generation, client/server desktop development tools. Imagine that you can deploy the results to large, and variable, user populations and it all still works... That sounds new! It also sounds like the new (8.1) version of WebLogic Platform. The Workshop tool allows you to do exactly that. But make no mistake, this goes much deeper than a tool. Anybody can host a load of windows in a single environment and produce a "developer tool portal," but that's not where the productivity comes from.

Remember, the productivity came from the assembly patterns and components, not from having a "universal hammer" that screws while it nails while it drills. The depth, and the value, is in the runtime framework that underpins the development environment. It applies sensible patterns and standards across the board, allowing you to treat low-level considerations like transactions as laws-of-physics rather than matters of pressing concern.

It Screws While It Nails, While It Drills!
So how does this relate to transactions directly? Well, in the process designer you draw your business process. If it's a long-running process, the framework will start a transaction at the start of the process and commit it whenever the process comes to a pause, to ensure the state is preserved. Probably what you'd want, right? For a process that doesn't pause it will start a transaction, do all the work, and simply commit it. Probably what you'd want too, right? But if the probabilities fail you, you can rubber band the transactions over your process steps, and tell the framework where they should start and end based on the physical realities of your system, still without a line of code explicitly referencing transactions. Oh, and by the way, when the framework assumptions don't apply to your particular use case you can drop back down to the J2EE APIs and the good old application server knowing that all the code will still end up deploying into the same environment.

Back to the Future
I remember my first column (WLDJ, Vol. 1, issue 1) talked about how transactions were assumed to be complex because all the two-tier developers never had to think about them, they just magically happened. And for the majority of applications, that's the way it should be. More strength to BEA's platform elbow in pushing them back into obscurity where they belong!