Github user mtaylor commented on the issue:
https://github.com/apache/activemq-artemis/pull/2392
> If the queue and topic with the same name/address were independent, then
clients explicitly selecting one or the other would not see each others
messages. Messages from a sender with capability 'queue' would not be seen by
subscribers with capability 'topic'. That is not how it behaves though.
Messages from all senders go to all bound queues.
Are you sure this is the case? If it does behave in this way then it's a
bug and we've broken JMS semantics.
>
> Two further separate points:
>
> (1) If the address already exists with multicast, and an amqp receiver
attaches with capability 'queue' (i.e. anycast), the address is not modified
and the queue is not created. The receiver is detached with an error that queue
support is not configured. (To me that is sensible behaviour). However a sender
requesting the same thing results in the address being changed and a queue
being created. That seems inconsistent at best.
Again, I don't think this is the case, but if you've experienced this with
AMQP then it's clearly a bug and we should get it fixed.
>
> (2) Once a sender has forced a pre-defined address to support anycast, a
receiver using a fully qualified queue name - myaddress::mysubscription - will
get an error unless they request 'topic' capability though surely this form of
address is entirely clear on what is required.
I don't think I get this one. The FQQN bit is meant to bypass all checks
and connect directly to the queue regardless of it's routing type.
---
