Conflicts

Inherent to any data synchronization routine is the possibility of a conflict. Conflicts arise when the same record (a record identified by the same key) is modified locally on different hosts. In such cases the synchronization routine has to make a decision about which data to pass into the new synchronized database and which data to ignore. Often, decisions are made on criteria like keeping the more current data (using time stamps) or by a user specified preference. For example, a particular user who makes changes to his appointment schedule on his PDA while his secretary makes changes to his appointment schedule on an office PC might want that the PDA data (updated by him) always overrides the PC data (updated by his secretary) in case there is a conflict.

Figure 2.1: Synchronization and Conflicts

An instance of conflict is shown in Figure 2.1. The record with key value $2$ in the database held by Host $A$ and Host $B$ has different data. It is up to the synchronization routine to decide what value the data field pointed by the key $2$ should be written to synchronized database, as shown in the figure. These decisions are resolved manually in case there is no rule set by the user about which data should be overwritten. We will not address conflict resolution issues here, a good discussion of which can be found in [16].