Hi all See http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.27.7611 In a nutshell, you cannot expect TCP streams to be error free, instead expect a packet to go wrong every few days; that's quite significant, particularly when talking about replication streams. And most if not all of us in the real world have encountered this. It's a serious case for having CRCs in the binlog (yes I know it's put in from Google patches to later versions in 5.5+ but I really want it in MariaDB, urgently!), but also as a potential consideration for other communications, including client/server. The replication streams are the most critical though, that's where stuff has been seen to break in practice - and that's when your lucky and the fail results in a syntax error - if it changes data rather than keywords, it can remain invisible. (in such cases you could have a replication slave copying a transaction "perfectly" except it has wrong data...) Regards, Arjen. -- Arjen Lentz, Exec.Director @ Open Query (http://openquery.com) Remote expertise & maintenance for MySQL/MariaDB server environments. Follow us at http://openquery.com/blog/ & http://twitter.com/openquery