Maybe I'm misunderstanding, but isn't the common approach to providing plug-in support in blueprint to use the "whiteboard pattern"?
On Wed, Feb 4, 2015 at 1:23 PM, Daniel Kulp <dk...@apache.org> wrote: > > Just to follow up on list about this to fully describe the issue and > potential solution… > > In the ActiveMQ schema that is used for both Blueprint and Spring, we use a > bunch of: > > <xsd:any namespace=“##other”/> > > in places where the user may want to add custom implementation of various > things. Plugins, transports, etc…. In Spring, the normal way this > would work is to use the Spring “bean” element to define a bean there. > However, blueprint does not have a top level “bean” element. The only > top-level element that blueprint defines is the “blueprint” element. Thus, > you cannot define a custom bean in those locations when using blueprint > unless you create a NamespaceHandler and register it and use a custom > namespaced element in those locations. Definitely more work than should be > necessary (and certainly more work than when using Spring). > > I just committed some changes to aries/blueprint to add “bean” and > “reference” elements to the aries “extensions” namespace handler. While > that’s not in the blueprint namespace itself, it should work. The users > would need to use something like: > > <broker …> > <plugins> > <bp:bean > xmlns:bp="http://aries.apache.org/blueprint/xmlns/blueprint-ext/v1.5.0"; > class="org.apache.activemq.plugin.StatisticsBrokerPlugin" /> > </plugins> > ….. > </broker> > > to get it to work. This will obviously require a new release of the aries > blueprint module and karaf releases and such. I’m going to do some more > testing and verification before pursuing that. No changes required for > ActiveMQ though. > > Any additional thoughts/ideas on this are more than welcome. > > > Dan > > > >> On Feb 4, 2015, at 9:32 AM, Hadrian Zbarcea <hzbar...@gmail.com> wrote: >> >> There is rather serious issue related to supporting blueprint in karaf. In >> case the jira [1] and it's relevance is not immediately obvious, I am >> pointing to it here to draw a bit of attention and choose among the two >> proposed solutions (or throw in other options). Lazy consensus defaults to >> the 2nd solution. Unfortunately, either solution requires a new bundle. >> >> Thanks, >> Hadrian >> >> [1] https://issues.apache.org/jira/browse/AMQ-5554 > > -- > Daniel Kulp > dk...@apache.org - http://dankulp.com/blog > Talend Community Coder - http://coders.talend.com >
