I fully agree with you, and would add one more point: In addition to SOA sucking conceptually, IBM's tools REALLY suck.

Where I work we were a wall-to-wall Microsoft shop until we got a new CIO who liked IBM, then bought what they had to sell -- including IBM WebSphere Process Server (WPS), possibly the worst platform I've ever used.

Not only was it a bad fit for our company -- but that didn't stop IBM Sales from telling us it was a cure for what ails us -- it's a REALLY poorly done product created by a bunch of obviously bright, but obviously clueless, researchers at IBM. Just a couple of the flaws:

There are few adapters for common external systems -- e.g. we needed to use an LDAP (Active Directory) as a source and target application (as opposed to a repository for WPS itself). No prebuilt adapters are available, so we had to try to write one from scratch in raw Java using the adapter API. What an unbelievable waste of time!

There's huge wasted effort on visual development where it's not useful, like the toy Java 'snippet' editor, wherein you drag little icons around to generate Java routines. A 'non-programmer' is meant to create these to perform little necessary tasks not included in the toolkit, like converting case or simple logic. Trouble is, you need to be a programmer to do this work, whether with icons or not. And if you can do that, you'll be faster with a regular Java IDE than with the visual editor.

On the other hand, when you could actually *use* a visual assist, it's nowhere in sight:

The flat file adapter -- dunno about you, but our shop has plenty of systems that communicate by reading and writing text files -- is not; it really only reads XML files. To read a text file of some arbitrary format -- say, pipe-delimited -- you get to write a parser in raw Java (not even the visual snippet editor!) Not a single visual tool to help interact with a sample file and generate this code, even tho tools like that abound, from little shareware utilities to powerful visual tools like DataWatch's Monarch or Pervasive's Data Integrator.

Finally, the gap between design-time and runtime environments is giant -- in one, you're in a cozy visual icon-driven environment (albeit with options tabs with billions of inscrutable choices). Then at runtime, if the slightest thing goes wrong, you're in the middle of a giant Java stack trace. A bit of effort on IBM's part to help *parse* that dump into something meaningful might help. The abstraction leaks like a sieve.

All in all, it's a crappy, expensive product that tries to solve problems we don't have, and doesn't do anything to solve problems we do have. Maybe it works for some people -- tho in searching for developers, we couldn't find any -- but definitely not for us. Run away.

So could you explain further those skills needed? Do traditional IT roles change? New roles?

"About half of all respondents admit they have less than 25% of SOA skills they deem necessary"

How do you quantify "skill"?

In other news: 95% of all statistics found to be completely useless.

At the risk of sounding like a skeptic... a lack of skills hasn't stopped software vendors from writing horrible software, why should it stop companies from implementing it?

I believe there is a perception that one can read a book, or get a certification, and be ready to "hit the IT world running"...

David, I have seen a similar study and agree that there is not only a lack of skills, there is still a real lack of understanding of what SOA means to an organization and the value that can be derived.
The good news is I see more organizations realizing value and willing to tell their story everyday. As that continues to happen we will see people and organizations make the investment, because it is an investment worth making. T. Harvey, www.xaware.org

"Don't waste time on projects that have no chance for success"

The reason that SOA hasn't taken off any more than it has is because it demands expertise in both IT and business processes in order to design effective systems. The number of individuals who understand both sides are few and far between. The best bet for companies that want to succeed with SOA is to engender better collaboration between Business and IT professionals. Neither side can succeed without the other.

I would say that the lack of "real" committment is killing SOA. Large organizations pay lip service to SOA and do not have processes in place for true sharing of information and increasing the collaboration among various departments and partners.

Once we have a business architecture supporting "real" SOA, you can find technical people willing to learn and implement SOA.

"... need to have a holistic understanding of all enterprise systems"

Holistic? Dear god please, not this word!

Most shops - big or small - are clueless about how things actually work. If that sort of comprehensive knowledge is required for SOA, its dead now. Only things that work despite "holistic" ignorance can survive.