One in three companies will screw up any project!

I agree. One in three is surprisingly low.

"The Standish Group surveyed 13,522 projects last year and showed that unqualified project successes are well below 50 percent, 34 percent to be exact. Out-and-out failures, defined as projects abandoned midstream, are at 15 percent."

IT Myth 5: Most IT projects fail

I think it's important to realize that just because ERPs are so widely used now doesn't mean that everyone's happy with them. There has continued to be a stream of criticism and discontent with the "silo" nature of ERPs, even if they do provide a level of cross-business-process integration (sic) that may not have existed before.

I would expect we'll see the same thing with SOA - that even if the original implementation goes wonderfully, over time the luster will wear off, and the management of a complex suite of services will become a major issue over time.

Great point. And, there is often positive PR on projects that are less than sucessful.

I also agree that SOA will change over time. If you look at the "SOA Software Stacks" they are adding everything under the sun.

The important thing is to think about software engineering differently - through reusable and compose-able components.

The reason why technologies like SOA etc. are now a buzz word in the industry is primarily because of integration problem that any end user face when running two or three applications.

ERP another big word for a common end user like me is something that is still seen as a problem solver as it tries to cater cross department business process via single application. Wow! I mean, I have finance, sales, purchase and everything with me if I buy only one application.

Issues crop up when there is a business process change in response to external factors or say business scale up. To incorporate new information and track down data evolution path, changes need to be made in existing applications. It is not easy. For my development vendor said a huge amount data now needs to be captured that has relations with other already captured information. And thus we need something to pull up all former information into a stage where the new data can be merged and etc. etc. I say yes, ok, so what as an end user I need to do....

Go for technologies like SOA, ESB to integrate my applications or better have a DW put into place. But will this solve the main problem. How can on earth we pre define at present what we May need to capture in future?

Is it possible to have some fundamental unit that can not only store and process data but can handle data as it evolves over time and space. Can we have a global data verse that is application independent? I do agree with Mike that almost all implementation will fail in SOA. The reason is we are trying to see what future information could be and trying to track it down Now.

The problem of data changing over time and the requirement to present a historical view of the data has been a data warehouse issue for a long time and there are tools to help. There are also tools that can present federated data as services. These entity level services can be consumed by business services and composed into a business process. The tools are there but getting the design right is the trick. You have to design for change, for exampling supporting versioning, and evolve SOA over time.