Joe McKendrick asks: "Who buys the first round of SOA?". In many cases, the first round is actually not explicitly billed as SOA. Check out the Case Studies on the Vordel website. These all describe projects which solved a particular problem, such as linking a Europe-wide transportation network into a centralized SAP XI system, so that shipment information can be seen in realtime or securely exposing tax information via Web Services. In many cases, the goal was a specific benefit to be realized using new XML-based technologies, but not a case of one department saying "Let's build an SOA and see what happens".
In each case study, the challenges of performance and security had to be overcome if XML-based technologies were to be used. But once these challenges are overcome, using XML Gateways, other departments can use the same (or similar) infrastructure for their own projects. This can be called an SOA. But the key point is that successful SOA initiatives start with real life problems which are solved, not by airlifting infrastructure into place and hoping for the best (a point echoed by the 451 Group's recent SOA report).