>Charles Hannum mentioned to be in passing some months ago that an implied 
>seek in disksort() might be appropriate (when disksort() rewinds to the 
>beginning of the queue).  However, this too would skew the metrics of 
>disks that do independent seeks.

Devices that support tagged queueing shouldn't use disksort.  I don't know
how you can properly determine the number of seeks for ZBR disks unless
the SCSI spec provides this type of statistic in one of the mode pages.

--
Justin T. Gibbs
===========================================
  FreeBSD: Turning PCs into workstations
===========================================