SOA, is that clear for everybody?

SOA story is a nice one! It was at the top of CIO challenges for 2007, and reemphasised in recent CeBit, ...

To consider the best way to develop, maintain application is to design them as independant LEGO buiding blocks, exchanging services is a nice creative dream.
But...
picture, in large companies, of application portfolio shows the real mess:
heterogeneous applications, different programming language, weight of history, big amount of usable application, only a small part being used and a smaller part really useful (using Renault CIO typology)

In that kind of case, SOA is more a modern (desperate?) programming way to try to make these old (and new) applications communicate, creating a layer of web services (communication layer using web xml protocols...).

OK, let's imagine it will work (in some years...)

Will that solve the "data base" problem, that created this heterogeneity? Every silo in the company has his own view (about what is client, what is a map, what is a piece of material) and therefore about the meaning, the way it must be coded, the information that must be inside....

Sure, a way to avoid all that is to move progressively to an ERP, insuring communication between silos but nobody can suddenly stop the past and move to a new way. And ERPs have got their own "philosophy of life"...

On that:

- is SOA DOA ? (Dead on arrival!)
- Good SOA synthesis by IBM (clear enough)
- SOA governance is a must, teams are key...