tkaemming
commented
9 years ago This is part of the replication branch as noted in the updated documentation for the relay configuration, so closing under the assumption that this will be merged in the future (and under the precedent of #22, #23, etc.)
Having to define a separate entry point for each handler makes it unnecessarily difficult to create a new handler implementation. This will also standardize the API that will be used for the replication processes.
(Also, "stream backend" or "log backend" is probably a better term than "handler", as noted in #4 — especially since the publisher is very coupled to the idea of a transaction log now.)