Re: Intermediaries - various cases

I think this is the second case.  The client is sending a request to a 
"Publishing Service", which has an implementation that forward the request 
to a bunch of subscribers.

I consider "intermediaries" are message pre/post-processing nodes along the 
path of from the initial sender and the ultimate receiver.  And 
"intermediaries" can only modify the route in a restricted way.  The 
initial sender MUST specify the destination and optionally a list of 
intermediaries (as well as which intermediaries it delegates authority to 
modify his route).

Rgds, Ricky

At 02:37 PM 9/26/2002 -0700, Ugo Corda wrote:

>The discussion about intermediaries near the end of today's conf call 
>reminded me of a couple of cases I have been thinking about in the context 
>of queuing infrastructures.
>
>Suppose I have a store-and-forward node to which I send SOAP messages to 
>be forwarded to particular destinations. I specify the destination in an 
>extension header, and I send the SOAP message to the store-and-forward 
>node. The store-and-forward node receives my SOAP message and sends it to 
>the final destination (possibly later, if the final destination is not 
>available right away). It looks like in this case the Service Requester 
>(according to our architecture diagram) is the node issuing the initial 
>SOAP message, and the Service Provider is the final destination node. So 
>the store-and-forward node must be the Intermediary.
>
>But now let's think about a case where I have a publish-and-subscribe 
>node. I send a SOAP message to that node with the intent that my message 
>be distributed to any subscriber (I don't even know the address of those 
>subscribers). So I send my SOAP message to the publish-and-subscribe node 
>(the only node I am aware of) without specifying any final destination. In 
>this case, is the publish-and-subscribe node still an intermediary, and 
>are the destination nodes Service Providers? Or is the 
>publish-and-subscribe node the Service Provider, which engages in separate 
>interactions with the subscriber nodes? Or is there no single answer to 
>these questions, and it all depends on the logical view that I want to 
>apply to the scenario?
>
>Ugo