[
https://issues.apache.org/jira/browse/ARTEMIS-1906?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16517534#comment-16517534
]
Justin Bertram commented on ARTEMIS-1906:
-----------------------------------------
[~lionel.cons], I think the regex idea is a nonstarter as I don't see how they
could be hierarchical. It is important that the syntax be hierarchical to
support the idea of general and specific settings which can overlap. I'm still
deciding on which direction to go here between the two options I mentioned
previously.
> anycastPrefix and multicastPrefix should be kept in mapped addresses
> --------------------------------------------------------------------
>
> Key: ARTEMIS-1906
> URL: https://issues.apache.org/jira/browse/ARTEMIS-1906
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Reporter: Lionel Cons
> Priority: Major
>
> For STOMP (at least), {{anycastPrefix}} and {{multicastPrefix}} are used to
> map destinations to addresses and routing types.
> In order to have compatibility with other brokers such as ActiveMQ 5 or
> RabbitMQ, these prefixes must be set to {{anycastPrefix=/queue/}} and
> {{multicastPrefix=/topic/}}. With these settings, a destination like
> {{/queue/foo}} works like a JMS queue while {{/topic/foo}} works like a JMS
> topic. So far, so good.
> Unfortunately, the current implementation removes these prefixes when mapping
> to addresses. This means that the two (completely unrelated, see
> ARTEMIS-1794) destinations {{/queue/foo}} and {{/topic/foo}} map to the exact
> same address named {{foo}}. Messaging does work as expected because the same
> address can have multiple routing types but monitoring and security become
> problematic.
> Security is defined per address so the two destinations {{/queue/foo}} and
> {{/topic/foo}} cannot have different security settings.
> Similarly, some monitoring information is attached to the address objects so
> {{/queue/foo}} and {{/topic/foo}} cannot have different monitoring metrics.
> It seems all these problems would disappear if prefixes would be kept when
> mapping STOMP destinations to addresses.
> Note: similar problems might exist with other protocols but I've tested it
> only using STOMP.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
