A URL does not name an object uniquely; the URL plus the MIME header issued with the request uniquely identify an object. For example, a WWW server may return a text version of a postscript object if the client's browser is not able to view postscript. We believe that this capability is not used widely, and currently the cache does not insist that the request MIME headers match when a request hits the cache. However, the cache does record the MIME header used to fetch each object.