[Daedalus]

Originally Posted by Pat the Cat

Actual chips used NCR53C96. Observe closely that both rows of each SCSI connector has pins connected.

I see a number of lines connected. Perhaps they were convenient vias to route a ground trace from. Notice particularly how most of the pins you're talking about are on the bottom rows of the connector (the ground pins), and are routed straight to the decoupling capacitors for nearby ICs. But of course, you're a "board-level hack specialist" or something, so you already saw that.

Yet according to some, half of the 50 pin connector on SCSI 2 controlers were always grounded. Never connected.

There's an extremely important difference between not connected and grounded that you appear to miss there. Anyway, what I actually said was:

"half the conductors on the 50-pin SCSI-I and SCSI-II ribbon cables aren't active"

I didn't say they weren't connected. They're all ground except for one, and are used for shielding and isolation purposes on long ribbon cables, similar to ATA-66 cables.

No change at all from SCSI-1 to SCSI-2 cables.

That's right. The bus speed changed but the connector stayed the same except for the 16-bit SCSI-II *Wide*, which uses a totally different connector to accommodate the extra bits.

Not that SCSI ever formally defined connections.

True, but there are several common standards, otherwise you'd never get anything to talk to anything. 50-pin IDC ribbons are one of them that applies to SCSI-I and SCSI-II drives.

the moment somebody plugged in an external drive, the whole lot ran at the slower bus rate.

Only for people connecting 8-bit SCSI-I or SCSI-II devices to 16-bit SCSI-II *Wide* or SCSI-III buses, which have totally different connectors. Such setups are rare as standard since external connectors exist for these buses and so they are used instead. Adaptors are needed to convert the bus to 8-bit, so if you're buying such adaptors, you surely know what you're doing.

I have a SCSI-II setup here now. It runs at 10MB/s. If I connect and external drive, it runs at 10MB/s. Absolutely no effect on performance.

People making devices that could talk at higher speeds always had the devil's own time demonstrating that it actually did so unless plugged into a suitable controller.

Obviously, but they're pretty foolish if they're trying to demonstrate speed on an incapable controller. Just like how hard a time you'd have trying to demonstrate the speed of a USB 3 hard drive by plugging it into a USB 1 host. People understand that. SCSI isn't voodoo; it might be more complicated and confusing than IDE, but shouldn't be beyond anyone even remotely technical once the bus requirements of the device are specified.