I attach a document that describes NSI in WS terms. I would like to discuss it tomorrow on the call. I am thinking perhaps we should stick to WS terminology and concepts for the first draft at least in order to make progress. I have taken the terminology from reading WS doc. It turns out to be slightly different than what we were using before, mostly in emphasis. In particular, the distinction between agent and actor is a little fuzzier, with the interface being between requestor agent that wants the service for its owner and provider agent that supplies the service. Actors come in only when discussing complex interactions between entities in particular infrastructure implementations. I think this is helpful also in thinking about service plane and what it is, though this doc does not deal with that. John
Hi John and all, I think Web Services may be used for NSI implementation, but not mandatory. Therefore, I do not think the architecture document should strictly follow the WS terminology. But using defined WS terms for different meanings in the architecture doc may introduce confusion. So, understanding WS terms is very important and John’s doc is very helpful. I read the WS document John referred "http://www.w3.org/TR/ws-arch/". Here are some comments: - "Agent" is the right word to denote a service component which exchanges messages defined by NSI. In the previous discussions, we used "agent" to denote a module inside a component. But this should be changed (as John explained). - The word "actor" seems to be used rather carelessly in the WS document. It only appears in specific parts (MEP part and the appendix on Web Services Security). So I think one (or two) of the authors of the document used this term without careful consideration. In addition, "actor" has a specific meaning in SOAP. So I think it is better to avoid using this term. - The terms "provider" and "requester" explain the role of an agent (or an entity, but I think we can forget about entity here) with respect to a service. One agent can be a "requester agent" or a "provider agent" depending on the roles. This is like one company can be a "buying company" or a "selling company" in commercial transactions. One agent may be a provider agent of one service and requester agent of another service (which is provided by another provider) at a same time. - The term "identifier" is explained as "an identifier should be realized (as) a URI" in the WS document. We need to be careful to use the term "identifier". Regards, Tomohiro On Tue, 19 Jan 2010 14:58:45 -0500 John Vollbrecht <jrv@internet2.edu> wrote:
I attach a document that describes NSI in WS terms. I would like to discuss it tomorrow on the call. I am thinking perhaps we should stick to WS terminology and concepts for the first draft at least in order to make progress.
I have taken the terminology from reading WS doc. It turns out to be slightly different than what we were using before, mostly in emphasis. In particular, the distinction between agent and actor is a little fuzzier, with the interface being between requestor agent that wants the service for its owner and provider agent that supplies the service. Actors come in only when discussing complex interactions between entities in particular infrastructure implementations.
I think this is helpful also in thinking about service plane and what it is, though this doc does not deal with that.
John
participants (2)
-
John Vollbrecht
-
Tomohiro Kudoh