Just a few thoughts I feel I should add here:

>SCSI termination and drive position make a difference under MacOS, but it's
>possible to have a SCSI chain misconfigured and still have a working system
>with tolerably low error rates.

very true, unfortunately.  However, mis-termination can cause 
unpredictable 
data corruption.

>Under NetBSD and MkLinux, I have found the
>SCSI chain to be rather less forgiving of hardware problems, possibly due
>to the less mature (no criticism intended!) SCSI driver code.
>
>A cheesy ASCII art diagram of the way I understand a SCSI chain should be
>connected on a Mac: (ms = mass storage device)
>
>  *----[]------[]------[]------@------[]----*
>term   ms      ms      ms    ctrlr    ms   term
>         e x t e r n a l       i n t e r n a l
>
>Often the terminator is on the control board of the adjacent drive;
>sometimes it's a device attached to the connector of that drive.  Note that
>the chain should be terminated at the ends furthest from the controller; 

Actually, the chain must be terminated at each end, with no terminators 
in the middle 
of the chain AND there must be no stubs greater than 6 inches long.  [ A 
stub is like a 
TEE.  Each drive is a stub, but it is very short  ( less than 6").]  

Your drawing is quite good.

>if there are no external drives, no external terminator is required.  One, and
>only one terminator should reside on each chain.

The reason for this, is that there is typically a terminator built in to 
the motherboard on
one end of the external chain.  So it only needs one on the far end if 
cables are attached.
One of the tricky things about termination is determining where the 
terminators are.  It USED to be that you could see them on the drive.  
Now-a-days, however, they are hard to spot.  What you need to look for is 
a small "jumper plug" on the configuration pins which is labelled 'TE' ( 
which stands for termination enable ).  Typically, when the TE jumper is 
ON, the terminator is enabled on that drive, and when the TE jumper is 
OFF the termination is disabled.

>
>In addition, it is considered good practice to place the SCSI devices with
>the highest bandwidth closest to the controller.  Hard drives are faster
>than CD-ROMs, which are faster than some optical drives, e.g.  This keeps
>the cable length short for the fast devices, where it counts.

The order of the drives is not usually very important.  Each drive 'sees' 
the total 
length of the chain.  The performance difference due to a drive being 
closer is totally 
negligible.

>
>Hope this simplifies the SCSI voodoo a little.  By the way, I was having
>problems booting my IIci with 3 external devices connected (including one
>very old Apple/Quantum 80 HD).  Replacing the old drive with one of recent
>vintage eliminated the booting problems.  I suspect it was a little too
>noisy or possibly flaky to communicate reliably with the rest of the SCSI
>devices.

I think what actually happened here is that the newer drive had 'active 
termination' whereas the older drive had 'passive termination'.   Active 
termination is slightly better since the variation in "Term Power' can be 
eliminated and the impedance can be better matched.


Make sure to check that each scsi device has a UNIQUE scsi ID.  If any 
two devices on the same scsi bus have the same scsi ID, it is GUARANTEED 
that your system will not work AND you may lose some data.  Also, some of 
the older systems ( before SCSIManager 4.3 ) required unique scsi id's 
even on different scsi chains.