In some applications, a certain percentage of records are discarded because of bad or missing information. For example, in an application where information on hard copy forms is entered in a database, a certain percentage of these records must be discarded because they are incomplete. In other applications, some data might need “fixing” before it can be used. For example, a field in a table might be designed to hold a string of comma-separated numbers, but might instead contain non-comma separators, letters, or other mistakes.
When data is synchronized, all updated data is dispersed, regardless of whether it is useful to a particular application. Bad data in one database becomes bad data in multiple databases when synchronization occurs. Therefore, when building a replication system for your application, you might want to consider writing procedures that either purge or fix data before it is replicated to other systems.