Jon Udell has a nice piece on the power of combining simple things viaServices and Links. The absence of SOAP from the discussion is a quality in itself in this discussion on web service infrastructure. Pure publishing and open link space are the original selling points of the web and they remain what it does best.
As Udell shows, this does not mean that the web becomes non-machineable, even if the lack of typing of most of URL space may make the services fail a lot. But that's a feature not a bug for some kinds of information.

Working out from that link via this link there was sort of a webbed thread about the semantic web effort and how it relates to the current vogue of piggybacking infrastructure on top of the html web via clever parsing and abuse of semantic elements in HTML. My ability to read the thread appears to be derived directly from the two-way functions in MT that we don't use at classy.dk since the embarrasing lack of readers would be then be readily on display, so that's a nice indication of where URL space could go as a mechanism.

The conflict between heavy standards based strongly typed hard to deploy solutions and easy to do makeshift solutions is a classic and the many weblog formats clearly have been able to build tremendous momentum from good enough interfaces. The main threat to the semantic web efforts remains the need to build infrastructure before attaining value and the unclear value semantic web efforts provides to first mover implementers. And of course any successful application would have to expect the data to be broken a lot and still be valuable.