bptm send various scsi commands, for example to set the scsi reservation on the drive, and then to check when the tape is loaded, and ready to be used, for example 'test unit ready' / 'mode sense'.
Depending on OS, these commands may go via the pass thru driver (eg. sg on solaris).
Once the drive is ready, the data is passed from the NBU data buffer to the os which writes it to tape via the tape driver. Also, when NBU sends the data to te OS, it sends the blocksize that should be used (SIZE_DATTA_BUFFERS). The positioning of the srive is done via the OS, this is where you see things like MTSFS / MTBSF in the bptm log - obviously, NBU reequests the positioning, but the actual commands are sent via the OS tape driver.
After a fragment is written, and/ or end of backup, NBU carries out a position check, asking the drive what position it is at, and comparing this to where NBU has calculated the tape should be. For example, this is how NBU would detect a SCSI rewind, it can't detect it when it happens (as it's not writing to the drive) and neither can the OS.
At the end (I think) , NBU checks for any TAPEALERTS that the drive has logged, and these would be dispayed in the logs/ error report.
SImple as that.
Martin
