Eating the own soup

In a recent communication of the ACM article, C++ inventor Bjarne Stroustrup wrote about the relation between academia and industry and common mutual misconceptions. There was one sentence that particularly caught my attention. "Professors should be able to code". Stroustrup, who spent most of his time in the industrial business and entered academia only recently obviously has a different take on things than most people in academia. He went on to say that students have detailed knowledge about algorithms, but essentially do not really know how to write a bigger program without it having numerous bugs. I think he hit the spot with this comment. We must ask ourselves critically whether we are really able to implement the things we are so easily talking about. Are these things even remotely possible? Or do they just sound good on paper?

The service community -  which I know best - seems to suffer from what  can be called a "lacking  implementation attitude". When reviewing papers, I often found very nice concepts without any clue as to how to actually implement them. It's irritating to read a paper of seven pages that presents "solutions" for an automated adaption of a business process without finding a single implementation other than some pseudo code like "loadPolicy(policy); applyPolicy(); modifyComposition()". In these cases, I often wonder if the authors have ever even implemented a single service and deployed it in a production system. After all, we are still engineers (at least most of us) and if we present a concept we should be able to implement it, too.

Only one step away from this lack of implementation attitude is something that I would call Evidence-based Service-oriented Architectures, or eSOA (in the service communitiy we obviously need nice acronyms :-)). In a nutshell, if someone writes a paper, they should present simulations of the presented concepts that can be verified or falsified by others. The main challenge is obviously to provide an environment that allows to publish the simulation settings to allow others to re-run the experiment. However, on a technical level there might already exist at least a partial solution to this problem. The so-called Genesis environment provides some of the features that are needed to generate SOA testbeds and to run simulations. Of course, there is still a lot to be done, but you can expect papers which will tackle the open issues in the future ;-).

- Martin Treiber