The Boa kills its prey by coiling around it and suffocating it. SOA is destined to do the same as the oxygen and budgets may be gone after the mega, multi-year projects.
Most successful SOA programs to date - amazon.com, Guardian Life - have taken 4-5 years of sustained commitment.
Very few IT shops are that disciplined and now the big software vendors and
SIs are salivating and offering to help them. SAP
has been re-architecting for SOA for 4 years now and is nowhere near
done. And as it gets done, its customers will face multi-year rollouts.
I read an interview with Don Rippert, CTO of Accenture in Consulting Magazine about SOA. He says post-SOA his implementation effort on a leading software package will be one-fifth of what it was prior to re-architecting for SOA. Pretty impressive - but listen to his caveat "Once the standards are finalized, once people are using the full vision of SOA...". How many zeros and hours does that take to get to? (I have an invitation to Don to write a Real Deal Guest Column)
I saw this post from Dion Hinchcliffe about convergence between SOA and Web 2.0. Web 2.0 has been about "light" as against SOA heavy - Google's constrained development model, Jotspot's economics, mashups over weekends.
Till the industry is ready to show rapid results - some semblance of "light" - we need to put SOA back in the lab. Trying to put the wolf in Web 2.0 sheepskin does not change the fact it is a wolf with a ravaging appetite. Actually, a boa with an appetite.
Update: James Governor points to a SOA success at a mid-sized company. As James says - not your average IBM reference. But it should be.
Phil Wainewright makes a great point. SaaS vendors inherently understand SOA since they have had to develop custom versions for scaleable usage. But guess who is trying to sell their versions ...on-premise software vendors and systems integrators who really do not have that much real experience supporting web services.