It was suggested by Peter Goeghegan that Slony could eventually use a binary copy to transport the sl_log data. The binary copy protocol is using network byte order and since the schema of the sl_log_1/sl_log_2 tables within one cluster will always be identical, even across different PostgreSQL versions and architectures, this should actually create no problems. I have no idea how much (or little) of a performance improvement that may result in. But it is worth investigating.