If I read the spec right, Content-Version reflects the contents of the document. That is, if the same document is dished up twice it is supposed to have the same Content-Version value. This may be a silly question, but it's probably not the last time you'll hear it, so I might as well ask --- how does this differ from an HTTP/1.1 entity tag (as used in the Etag:, If-match:, If-none-match:, and If-range: headers)? (These are opaque tags which are intended to identify different variants of a resource for caching purposes. They come in two varieties, "strong" and "weak". A strong entity tag always denotes the exact same set of octets. A weak one may denote multiple versions --- in the colloquial sense --- with minor, semantically insignificant differences, which may still differ enough so, for instance, you can't mix byte-ranges of different ones willy-nilly). rst