Well that was already clear to me, but what I can't understand is why such a big changeset should fail in the first place? I mean we are all testing that stuff and of course it can fail sometimes, but mostly you will have to do manual changes on a system on which a changeset failed anyway. Even if it did fail, as you pointed out, I thought that the changes are executed within a "transaction" and are therefore rolled back if they fail. If DDL does not support a transaction, changes could be reversed, couldn't they?
Maybe I am missing the point here but having many changesets feels kind of wrong for me. It would really help me if you could give me an example scenario or point me to one, in which the many changeset approach was actually helpful.
Maybe I am missing the point here but having many changesets feels kind of wrong for me. It would really help me if you could give me an example scenario or point me to one, in which the many changeset approach was actually helpful.