But why burden the server with anything more than a reference to
an object? That would require a distributed namespace implemented
in Java. And if we could get beyond naming, a lot of cool facilities
would be possible such as replication, process migration, agents,
distributed queries, as well as your favorite list of facilities.
Why limit the Web to being a big "file system" when systems like
Lotus Notes or Microsoft Network tell us that there a better ways
of doing things and these involve ease of replication, storage,
scripting and ability to compose arbitrary chunks of content?
However it seems that even naming such a web-wide file system is
a substantial innovation for the web, not likely to happen soon.