Would you care to publish a benchmark and suite?
There is no benchmark. Simply I noticed that on my console and I was just wondering if there might be an obvious reason that I don't see....
I found some interesting info in the spec:
section 6.2 of the jms spec:
Since there is typically some latency in all pub/sub systems, the exact
messages seen by a subscriber may vary depending on how quickly a JMS
provider propagates the existence of a new subscriber and the length of time a
provider retains messages in transit.
For instance, some messages from a distant publisher may be missed because it
may take a second for the existence of a new subscriber to be propagated
system-wide. When a new subscriber is created, it may receive messages sent
earlier because a provider may still have them available.
JMS does not define the exact semantics that apply during the interval when a
pub/sub provider is adjusting to a new client. JMS semantics only apply once
the provider has reached a ?steady state? with respect to a new client.
There is zero logical difference between a Queue and a Topic in terms of latency.
The spec is just referring inherent latency in any asynchronous protocol
such as JMS, rather than a synchronous protocol, e.g. RPC.
You pay your money, you take your choice.