The identifier of the node to store the comments is build like that: urn:xmpp:comments:_<random UUID>__urn:xmpp:groupblog:<publisher> Why don't we use instead of a random UUID the main item identifier to which the comments are associated? This is also a UUID. Doing that will allow to guess the main item identifier just looking at the comment node identifier and vice versa.