Re: [wildfly-dev] Easing migration beyond the migration ops

On 8/12/15 4:54 AM, Tomaž Cerar wrote: Thinking about this more, I don't think it's worthwhile. First, doing this would make the subsystem go away, but the extension will still be registered, and the user will still need to remove it. This trick only does half the job. Until the user does the other half the API to re-add the subsystem is still there to confuse users (although we'd make sure any attempt to use it out of --admin-only fails) and there would still be WARN logging during boot. Second, we wouldn't add the subsystem, but until some change to the config file happens causing a write, the subsystem will still be in the xml. That's confusing if the user notices it and violates the fundamental principle that the config file accurately represents the current in-memory config. Finally, as I mentioned elsewhere in this thread, the cmp and jaxr subsystems also ship with an empty config, but they aren't good candidates for this treatment because with those the empty subsystem really means something. So there'd be different behavior between threads and those, for non-obvious reasons. -- Brian Stansberry Senior Principal Software Engineer JBoss by Red Hat