On Tue, 2011-01-18 at 11:13 -0500, Alan Conway wrote:
> 
> I don't see this as something that applies in deployment. You're never
> going to 
> deploy brokers without knowing in advance what port they're using, the
> broker's 
> host:port are the fundamental location information that clients have
> to know to 
> connect. I don't see any non-test use case where a client knows the
> broker's PID 
> but not the port. 


OK, I guess I was trying to sneak in a Discovery use-case, at least
local discovery, into this.

The point with that was that -- even if you know the broker ports ( and
whatever else similar ) in advance, how do you communicate that
knowledge to apps that start up after the brokers are running?

We should have a standard way of communicating that knowledge.

***
I must admit, though, that local-discovery would probably best be done
as a special case of Global Discovery, and that the final-config-echoing
solution you suggested here sounds the like the best bang for the least
buck for this limited "Easy Broker Info" idea.
***

When we do the (separate) Discovery feature it should include not only
brokers/ports/transports but also available clusters & something about
federations.

i.e. an app should be able to wake up knowing only the name of a cluster
it wants to connect to, and be able to easily discover the ssl and tcp
ports of that cluster.

Also it seems cleanest to have a solution that allows the ports
associated with clusters/feds to change over time.  New brokers get
added or subtracted.









---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:[email protected]

Reply via email to