All this while, I used to think SOA based on WS-* (see W3C and OASIS) would likely not scale because of the complexity involved in understanding and using the WS-* specs. Now I’m firmly convinced that SOA based on WS-* will never scale because the entire philosophy and thinking that is driving it is fundamentally flawed. Let me explain.
The idea behind SOA as a concept is loose-coupling, i.e. about a large body of applications that co-operate with each other to produce more powerful super-applications. If you look at the thinking driving SOA based on WS-*, it is all about tight-coupling, i.e. about a large body of applications that are extremely well co-ordinated to produce more powerful super-applications. Co-ordination instead of co-operation – therein lies the flawed thinking.
Co-operation is a notion that can be made to work on a large scale with few guiding principles/standards (the web is a wonderful example; others are the movie industry using DVDs, printer industry using letter size paper, etc). It is not easy to achieve, but it is feasible. Co-operation allows for individuality while enabling a shared goal to be achieved.
On the other hand, co-ordination on a large scale is practically impossible to achieve (an army like the US army, the New York Philharmonic Orchestra are some examples of co-ordination). Co-ordination demands extreme discipline and control. Individuals do not matter, only the common goal does.
Everything about SOA based on WS-* tries to achieve co-ordination (the specs themselves and all the grand talk about ESBs, governance, lifecycle management, meta data repositories, etc).
Co-ordination is extremely appealing because it promises ultimate efficiency. This is probably why enterprises are taken in by the SOA based on WS-* sell. With enough time, when it becomes apparent that they’re chasing a fallacy, people will come to their senses and do something much simpler to achieve co-operation that may be messy and not so efficient, but it sure as hell will work.