"adrian(a)jboss.org" wrote : | Only if TestJMSLocalProvider points at a plain ConnectionFactory, i.e. | one that doesn't implement XAConnectionFactory. | | If the jms provider implements XA then don't use a local connection manager | to manage transactions. You end up with the weaker Last Resource Gambit | (i.e. the transaction observer pattern rather than the transaction participant pattern). | | But I'm just repeating what I've already said. | | I want a real use case not trying to support some broken/stupid configuration. IMO if the user wants to configure the connection manager for local only, then that's what should happen. The user may have reasons for doing this, like they specifically don't want to use XA, e.g. if there is a known bug in the provider's XA implementation and they need to roll back to local tx as a workaround. So I won't necessarily consider this to be broken configuration. What shouldn't happen is the system silently fails in subtle ways, as it does now. Inferring whether the user wants to use XA on the basis of whether the CF implements the XA CF interfaces is not valid - there's nothing in the spec that says the the same class can't implement both the XA and non XA interfaces. *** Tim dons hard hat in preparation for brocking View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4142850#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...