[ https://issues.jboss.org/browse/WFLY-11166?page=com.atlassian.jira.plugin... ] Brian Stansberry commented on WFLY-11166: ----------------------------------------- [~dmlloyd] So all the existing Service injection should be ripped out? That's getting to be a bit beyond what I'm up for doing. But OTOH perhaps I should just "Won't Fix" this and instead add capability requirements for the existing org.wildfly.transactions. There were really two things that sparked my interest in this: 1) Subsystems depending on having a local provider but the management model (and hence galleon too) not having a clue. So you could remove the tx subsystem in admin-only mode and it would be accepted, but then you start in normal mode and boom. So just requiring org.wildfly.transactions or perhaps a more precisely named org.wildfly.transactions.local-provider fixes that. 2) Avoiding classloading dependencies on org.jboss.as.transactions, which itself has a lot transitive dependencies, so possible source of big ball of mud. But investigation since then shows there's very little benefit from not depending on org.jboss.as.transactions (at least not if your feature really requires a local provider). The key thing is to avoid unnecessary hard dependency chains that propagate out from org.jboss.as.transactions, e.g. WFLY-11195. -- This message was sent by Atlassian Jira (v7.12.1#712002)