People usually miss the references that should be encoded within REST responses, The hypermedia part. You should be able to call /user/ and get a list of users whith links to the specific users. This link will be in the correct http-format which means you can discover the api from one or a few starting points. This gets even more obvious when you try to fetch a sub-resource or call a "method" on a resource. Those links are encoded within the resource itself. That means when you get a user record it may contain links to things you can do with it (add an address or an extra email, get specific details etc). With SOAP it is only naming and there is very little recommendations about that.
I guess what I object to is this concept of discoverability. No matter what, you need documentation to discover these initial endpoints, so why not document the whole API? SOAP had mechanical discover ability with WSDLs but it didn't really work. The reason REST won was because it was simple for humans to understand, which means mostly that the documentation is simple to read. As long as the URLs follow some simple consistent pattern that developers can implement without convoluted effort it should be fine, no need to follow some rigid philosophy. If the documentation is missing some key actions I doubt most developers are going to go poking around hoping to discover undocumented endpoints, they'll probably end up contacting the developer to tell them they're API is broken.
