> > That's always the case. 
> 
> I don't agree.  The IDE controller is much more analogous to a completely
> stupid SCSI adapter that requires host intervention every time it does
> *anything* with any command.  It's the problem that the hardware doesn't
> have a local, managed queue of commands that makes the interrupt and latency
> problems so severe.

Oh, well, I meant that there's always some kind of overhead, whether on the
local HBA or on the target, for managing multiple commands.