i thought the algorithm for distributed transactions was a 3 phase commit? how is this faster that replication? have you done a proof of concept to prove this is faster before you develop it? run any tests? you have have replication update your remote database on the fly and you do not have to issue a refresh command. i dont get it. all i see you doing here is user hits a button. this button leads to data add/change/delete from a table. you need this data immediately pushed to a remote DB. so at the same time you execute a package to push it to the remote DB. how is this faster than using the built in stuff to update the remote DB on the fly? -------------- Original message -------------- > Hi All, > > Environment: > We have 2 databases (both are active) physically located in different -- //www.freelists.org/webpage/oracle-l