Got your mail and sent you a response.
Do you have any idea how fast the filestore code [could be]/[will be] as compared to a Postgres dbase-backed mail store, especially in a large scale implementation?
For larger email providers, it seems like a SAN would be employed (whether with a filestore or a db store). Does the database add anything beneficial in such a scenario (for typical mail operations). Or, would it make sense to employ some kind of hybrid (mail messages on the FileSys, Calendaring/User data in the DB)?
View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=3990710#3990710
Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=3990710
The recent changes on HA, especially adding the concept of Replicator, added a dependency of ClusteredPostOffice on ConnectionFactoryJNDIMapper.
The ConnectionFActoryJNDIMapper is started by ServerPeer what causes a circular dependency between ServerPeer and ClusteredPostOffice.
The right way to fix this would be having an injected dependency (it would be cool to have PojoContainer here). Right now I'm changing the code to make ConnectinFActoryJNDIMapper to get and register replicator lazyli... but that's not the right thing (just one thing to have it running now).
The right change would be to have ClusterPostOffice injecting replicator on ConnectionFActoryJNDIMapper. If we don't go to PojoContainer soon we should think about doing that. If Tim Fox is cool about that I will make this change next week until we have MicroContainer integrated.
View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=3990684#3990684
Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=3990684