The original idea was that you had URLs that told you where to find a thing, and URNs that identified a particular thing regardless of where it appeared. These used a unified schema (called URIs) because when you’re telling somebody else about a thing, you want to be able to refer to it by location (“the red building in the next block”) or by identity (“the main post office”). Presumably, there would be services that could take a name and tell you where to find it; not a bad assumption as every library in the world has such a system for its own collection.

While this looks good on paper, in practice URLs were relatively stable in the early days of the internet and so they turned into de facto names before lots of effort was put into making URNs work. Now, we’re struggling with the issues they originally foresaw, but weren’t able to follow through with a working implementation.

(For more details, see https://danielmiessler.com/study/url-uri/ where I got much of this information)