Every table that is to be synchronized must have a primary key, and for each synchronized table the primary key must be unique across all synchronized databases. The values of primary keys should not be updated.

It is often convenient to use a single column as the primary key for tables. For example, each customer should be assigned a unique identification value. If all the sales representatives work in an environment where they can maintain a direct connection to the database, assigning these numbers is easily accomplished. Whenever a new customer is inserted into the customer table, automatically add a new primary key value that is greater than the last value.

In a disconnected environment, assigning unique values for primary keys when new rows are inserted is not as easy. When a sales representative adds a new customer, she is doing so to a remote copy of the Customer table. You must prevent other sales representatives, working on other copies of the Customer table, from using the same customer identification value.

This section describes the following ways to solve the problem of how to generate unique primary keys across all synchronized databases:

• Composite keys
• UUIDs
• Using GLOBAL AUTOINCREMENT
• Primary key pools