Snapshot downloads   If you are performing snapshot downloads, then the initial download file just needs to set the generation number. You may choose to include an initial snapshot of the data in this file, but since each snapshot download contains all the data and does not depend on previous downloads, this is not required.

For snapshot downloads, using the -bg option is straightforward. Just specify -bg in the dbmlsync command line when you create the download file. You can use the same script version to create the initial download file as you use for subsequent download files.

Timestamp-based downloads   If you are performing timestamp-based downloads, then the initial download must set the generation number on the remote database and include a snapshot of the data. With timestamp-based downloads, each download builds on previous ones. Each download file contains a last download timestamp. All rows changed on the consolidated database after the file's last download timestamp are included in the file. To apply a file, a remote database must already have received all the changes that occurred before the file's last download timestamp. This is confirmed by checking that the file's last download timestamp is greater than or equal to the remote database's last download timestamp (the time up to which the remote database has received all changes from the consolidated database).

Before a remote database can apply its first normal download file, it must receive all data changed before that file's last download timestamp and after January 1, 1900. The initial download file created with the -bg option must contain this data. The easiest way to select this data is to create a separate script version that uses the same download_cursor's as your normal file-based synchronization script version but does not have a modify_last_download_timestamp script. If no modify_last_download_timestamp script is defined, then the last download timestamp for a file-based download defaults to January 1, 1900.