Github user michaelandrepearce commented on the issue:
https://github.com/apache/activemq-artemis/pull/1695
@andytaylor the issue i see is that the defaults should err on the side of
safety imo.
Having a client not throw any exception and continue blindly is dangerous
(even if its miss-config). Just because something is like it is, doesn't always
been it should be.
I get the performance thing, but then thats something someone should make a
conscious decision to enable at the expense of exceptions.
Just a bit of background, we migrated a flow (it was using JMS api, so
switch is just in the CF) from another vendor's broker to Artemis, and released
to prod.
The broker did alert, but the operations team operating the broker saw it
as a client issue. and as was a client issue they believed the client team
would also get an error, have been alerted and would have resolved.
Because the client team had no error, it was not until quite a period later
it was discovered and external reports had to alert us to the issue.
We re-tested the scenario with our other brokers, and they would have
alerted client side so this is why this assumption is had.
If we look to other people migrating I'm fairly sure the other brokers we
use, are quite well used, and as such other users would make similar
assumptions and expectation that the client would still error by default,
unless they toggled something for performance on purpose.
---
