Yes, but it still presupposed on a specific location. Alan Kay (together with many other experts on system design) is a proponent of late binding. To do effective late binding, we should be able to route requests to the correct location in a distributed fashion. With a (regular) URL we make a weird amalgam of DNS (with IP) and path of the object. As we've learned from the transition from XML-RPC/SOAP, unnecessary complexity removes value of the system. I believe Alan argues that URLs introduce unnecessary complexity which can be tackled by lower OSI layers. However, to make that work, we need globally unique identifiers and routing, using IPv6. We want names (and/or identifiers) to be the fundamental element of our systems, instead of resource locators.

In other words, you want to be able to walk into any library and use ISBN as the underlying identifiers, instead of the Dewey system, which is not broadly applicable. You might get the Dewey number from the ISBN and vice versa, but the ISBN resource assigned to a book resource is the fundamental linking pin.

So, when we link from an article, we want to be able to say: "it is this specific article I am linking to, you figure out how to get access to it." This will be much more durable than an ephemeral URL.