“A Snapshot of Public Web Services” (warning PDF) appeared in the March 2005 issue of the SIGMOD Record. It discusses how many of the publicly available ‘web services’ are really just data retrieval services or search services. Available services also tend to be fairly poorly documented (fewer than 10 words) and in more cases than not, contain malformed WSDL. In fact, many of the services reported on tend to be more at the ‘Hello, World’ end of the spectrum — their existence in the directory more out of experimentation than utility.
It should be noted that the services in question were trawled out of xmethods.com and the like, so they don’t represent the services or APIs that sites like Flickr, Google, Amazon or others expose.
I believe that it would be very valuable to have an inventory/registry of proper/richer services that are available on the Internet. This would certainly facilitate the growing trend of people wiring up one sites functionality with another sites services to create something altogether different. Such a site could enforce some level of validation and even provide semantic web capabilities such as OWL-S description and composition facitlities to aid the systematic discovery and usage of said services.