Most sort order problems in Replication Server are a result of different sort orders being used in different servers throughout your system.
You receive a Data Server Interface (DSI) error when trying to insert or update a record into a table with a unique index on a character column. Under the primary data server’s sort order, the value is unique; but under the replicate data server’s sort order, it is not unique. For example, “lvis” and “LVIS” are considered distinct under a binary sort order but are considered equal under the nocase sort order.
A drop subscription with purge command does not completely remove the rows at the replicate database. A query to select records to purge misses some records under the replicate data server’s sort order.
Replicated data is not what the replicate site expects.
When this occurs, a subscription probably has one or more character clauses. Updates during the lifetime of a subscription are processed with the primary Replication Server’s sort order, leading to unexpected semantics at the replicate data server if the sort orders differ.
Correct these problems by editing the Replication Server configuration file to use the same sort order and, if necessary, by running sybinit to make the primary data server and replicate data server use this sort order as well.