We should make sure the spec has an agreed definition of the term "subscription" and uses it consistently,
There is currently a definition in WD-12.. "The request made by a client to receive Payload Messages published by Clients. The Subscription contains a Topic Name which can include wild card characters so that it may match many Topic Names."
This seems to make "Subscription" equvialent to "Subscribe" Control Packet,, whereas we probably want something that refers to the stateful object that is maintained by the server on behalf of the client. Also, even if we wanted it to mean that, a Subscribe Packet can contain more than one Topic Name, so the second sentence would be incorrect anyway