I was re-reading Graham Murray's comment to a post I made earlier about Web services and the end of integration costs. It got me thinking...
I don't see the realization of SOA as killing off integration nearly as much as I see it opening the doors for future work. As Graham noted, just because something is wrapped in a Web service doesn't mean it is ready for consumption by everyone and their dog. He makes the point that a legacy system may be reporting in pounds when another system may need the info in euros.
How easy would it be to take a webservice that reported info in pounds and expose a new web service that reported info in euros. Now both could publically be available. If your conversion rate was available via a web service it would be even easier.
The key is that it all becomes easier but there will also be money to be made. Imagine you are a corporate manager of some sort and you are getting these hard copy reports off the mainframe printer that tell you sales in pounds. You grumble thinking that you wish you had the info in euros but realize that it would cost you an arm and a leg to get someone to write the conversion capability for you--particularly because it is getting harder and harder to dig up cobal programmers anymore.
Now suppose it is available in a Web service. Hey, you can get that report done really easily and at little cost so you are willing to spend the money. Particularly when you realize that the data will now be able to be used by any number of other apps or reports.
It will not be the end of integration...in fact it will be a launching pad for a lot more integration. There will just be very little spent on things like converting strings on one platform to strings on another platform. Instead people will be able to focus on the benefits of integrating in a trivially cheap integration environment.