[JXTA discuss] JXTA propagation change of behavior

While observing network performance, stability and measuring the number 
of messages received by an idle JXME proxyless peer, I have stumbled 
upon a bug <http://platform.jxta.org/issues/show_bug.cgi?id=1404> in the 
WirePipe which re-enforced my opinion of group propagation, and led me 
to deploy a point patch on the public net, which has been running for 
the past six days, and the results have been very pleasing.  To give a 
little background :

Currently all peers (edge and rendezvous peers alike) have the ability 
(without discretion) to propagate a message to all members of a group 
(mainly JXTA 1.0 legacy), which in turn puts a big burden on all members 
of the group (especially rendezvous and relay peers of a group).  A much 
better alternative is to use the ResolverService, as it provides the 
application/service full control over stopping or continual of a 
calculated (based on the JXTA network topology, and a peer's 
designation) walk through the network (as a matter of fact, several of 
the core services use the resolver for such purpose (see Discovery, 
PipeResolver, PeerInfo, EndpointRouter), whereas a propagated message 
expires when the ttl is exhasuted, or it has been processed by all  peers.

As a result Rendezvous.propagate() will be deprecated, and the it's 
behavior will be altered on the wire, such that a propagated message is 
delivered to the destination service or application, and re-propagated 
no further. If your application relies on rendezvous.propagate(), you 
are forewarned,  you need to rely on alternate services such as the 
resolver, or propagated pipe.   If you rely on rendezvous.propagate(), 
and don't see how an alternate method maybe implemented, then please 
forward your concerns and questions to this list.

Regards,
Mohamed

-- 
http://blogs.sun.com/roller/page/hamada
http://weblogs.java.net/blog/hamada