[rabbitmq-discuss] Exactly Once Delivery

Sat Aug 7 21:47:00 BST 2010

Matthias Radestock wrote:
> At that point the cloud messaging service has two copies of the same
> message in different locations. For the duplication to be detected, some
> information needs to flow between the two locations. Which is expensive.

Aha! I see. Thank you. I had the *publishers* being in different locations in
my mind. Which is of course a different problem.

In the context in which you originally posted the example ("using
publisher-supplied message ids for de-duping simply does not scale") I suspect
that the mention of publisher-supplied message IDs is not relevant: it's the
maintainance of any kind of dedup buffer at all in more than one place at the
same time that's not scalable.

Isn't the fundamental state-synchronisation-between-server-nodes problem
independent of the choice of message ID?

Tony