The real problem with Xanadu is that it's based on a fundamentally flawed concept -- knowable provenance. The problem is one of having an overly literal and simplistic view of what a document is (indeed, this problem is embedded in the announcement page itself). Such and such is linked to THE ORIGINAL sourcedoc. The problem is that with almost any significant document provenance is ambiguous and unknowable, which turns anything of significance into an amiguous mare's nest of noise links.

Take for a simple example "Macbeth". We don't have the "original", we don't know for certain which version derives from which version, and we don't know for certain who is the source of any given change. Yet Nelson uses a shakespeare play as one of his examples blithely ignoring the problem of provenance throughout. Now imagine a really thorny issue like the Bible or a particle physics paper.

Imagine if every document in the world were created and maintained using the same bug free version of word with change tracking on. Any quotations or abstractions were carefully done using OLE embedding. No-one ever misquoted someone else from memory, or accidentally deleted a file and then typed it back in from an old print out. And all the edits acted as links. Ok it still wouldn't work. Welcome to xanadu.

Comparing the web unfavorably to xanadu is like DOS fans decrying the mac because they could do so much more from the command line, except that DOS actually worked at the time the Mac came out.