Malte,
Assuming that the callback implementation will go ahead, I would re-iterate that the callback seems to be dispatching the message to a particular handler (its owner) - and therefore I think it is best to have a predetermined scheme for addressing the owner. Which then results in only one implementation being called. Unless, of course, you can think of a plausible use case in which several message types need to be handled by a mixture of several handlers.
As Gustaf points out, the principle function of the callback is a message dispatcher and can be thought of as similar to the request processor. The scheme for decoding the message (URL or email address/subject/body) should be the same regardless of the handler.
This contrasts with the 'DNS blacklisted URLs' handler example above in which the callback implementation is acting not as a dispatcher, but as a filter.
