The url alone is not sufficient to identify a unique unit of web content. There are lots of cases where the same url returns different things depending on the request headers. We should include at least some of them in the identifier.
Fundamental change with much trickiness involved! How feasible?