Reading a piece on El Reg about London Underground and Vista highlighted brilliantly something that has been batted around internally at work.
For Web 2.0 to really deliver the hard work is getting the existing systems ready, not in the flashy GUI. In the demo above the London "frozen points on a railway line that is always >10m" Underground knocked up a very funky demo using 3D rendering and all those flashy Vista collaborative and display features. The hard work they say had already been done in exposing systems in a way which could actually be consumed like that.
The point here is that not only do SOA and Web 2.0 work well together its actually really hard to see how you can have enterprise grade Web 2.0 without changing the way you deliver your existing IT. Please note here that again I'm not talking about WS-* v REST as those are just implementation technology decisions. What I'm talking about is creating an existing IT estate that can be easily consumed by "Mashup" or dynamic applications.
This shift requires a lot more than just lobbing Web Services onto systems as doing that would rapidly bring most mainframes to a grinding halt. So what you've got to do is make the estate function in that way from the perspective of the consumers this means that people might think that they are accessing the mainframe directly but in fact there is a sophisticated cache in front of it. This means presenting services in a way that is sensible to be consumed, and yet again this isn't about technology, its about taking a consumers view of service description and thinking about the exposure of the estate both in terms of the services and the Virtual Services consumed externally. The key here is that a lot of future value for organisations is going to be based around that form of external collaboration with suppliers, customers and partners. Enabling that is a goal of both SOA (thinking), SOA (technology), Web 2.0 (technology) and Web 2.0 (Kool-aid drinking PPT jockeys). This won't be done in a green field environment it will be done based on the existing applications, ERPs and mainframes that the organisation has today and which were never intended to work in that way.
So while there are some cracking Web 2.0 apps out there using existing technologies its going to be a massive change for existing IT enterprises to cope with this switch from I know the user now and I know what they are going to do applications to I'll know the user when they call and I'll know what they want when they ask. This is the key behind the switch from user based systems to participation based systems and the key challenge for Web 2.0 delivery.
If organisations don't embrace these thought and management changes, and then (N.B. and then) have the technology to deliver the change then all Web 2.0 will be is a fancy dress and some lipstick on a butt ugly pig.
SOA doesn't need Web 2.0, but its looking like being the best interaction model for future systems.
Web 2.0 does need SOA if its going to help enterprises deliver external value.