<Well, maybe -- but if you're doing programmed I/O, I expect the bottleneck
<will be in how fast your CPU can sling bits.  The performance of the SCSI
<on the desktop VAXen, in PIO mode, is so horrific as to make a PIO-only
<IDE controller seem well-nigh worthless...

PIO for the SCSI controller is very differint from simple PIO to a port.
Look at the SCSI code that does it, then take everything out that is 
specific to the SCSI driver and device.  VAXen do PIO real fast, PDP-11s
aren't to shabby either.  VAX CPUs however have a high interrupt overhead
that makes them slow for IO.  PDP-11s do not suffer from that.

<The wire protocol for the "UltraDMA" and "DMA" IDE modes is evidently 
<different.  I don't know how, Manuel Bouyer could probably tell you.  The
<old-fashioned "DMA IDE controllers" used ISA or EISA motherboard DMA and
<just used it to push bytes into the FIFO once you got things started --
<that worked with any drive, the new modes do _not_.

Can't be much different as you can mix them on the same two plug wire.
Your limited to the slower of the two however.  For Qbus DMA is all that 
is wanted or needed, only to get away from PIO.

Allison