RE: Federation vs P2P discovery

Paul,
 You wrote:
   >Katia, I think, disagreed and introduced the notion of peer-to-peer
(P2P) 
   >discovery.  Was it that a choreography would not apply in P2P
discovery?

My introduction of p2p discovery was to say that in the document we did not
have p2p as a discovery method and I volunteered to add it.

>Katia also seemed to distinguish between several exchanges to a UDDI 
>registry (find business, get business, get service, get binding), versus a 
>sequence of exchanges that cross from one registry to another (perhaps, 
>from peer to peer?).  Is the point that a sequence within a registry would 
>not require a choreography, but exchanges across boundaries would merit a 
>choreography?

The definition of choreography in the current draft of the WSA document
says: "A choreography defines the sequence and conditions under which
multiple cooperating independent Web services exchange information in order
to achieve some useful function".
If we adhere to this definition, then several exchanges to a UDDI registry
would not require a choreography. Now, whether federation of registries or
p2p based discovery (eg a Gnutella like propagation of discovery messages)
requires a choreography is debatable. For example, in a federation of
registries a request for providers with particular desired capabilities may
be forwarded from one registry to another if the first registry does not
have any providers of the desired capabilities.  This forwarding is not
however an "exchange of information" in the sense implied by the
choreography definition; also in the case of federated registries, the
choreography for a particular request would vary depending on whether the
requester happens to ask first a registry that had the provider or another
registry that, because it does not have the provider must propagate the
request. Same notion for p2p (since again here we have forwarding). [And, of
course here the specter of intermediaries and the long discussions that the
group has struggled with raise their ugly head:-)]So, whether federated or
p2p discovery require choreography also depends on whether we (as a group)
agree that the notion of choreography (for the achievement of some
particular function) should be static i.e. the same for the same function. 

 Cheers,
 --Katia

-----Original Message-----
From: www-ws-arch-request@w3.org [mailto:www-ws-arch-request@w3.org] On
Behalf Of Paul Denning
Sent: Thursday, December 11, 2003 5:55 PM
To: www-ws-arch@w3.org
Subject: Federation vs P2P discovery


On the WSAWG telecon on 2003-12-11, I discussed some thoughts about 
modeling discovery as being described by a choreography.

Katia, I think, disagreed and introduced the notion of peer-to-peer (P2P) 
discovery.  Was it that a choreography would not apply in P2P discovery?

Katia also seemed to distinguish between several exchanges to a UDDI 
registry (find business, get business, get service, get binding), versus a 
sequence of exchanges that cross from one registry to another (perhaps, 
from peer to peer?).  Is the point that a sequence within a registry would 
not require a choreography, but exchanges across boundaries would merit a 
choreography?

Please clarify the points you were raising.

Paul

Received on Thursday, 11 December 2003 20:45:16 UTC