Subject: cc:Mail Link to SMTP Undeliverable Message
To: None <port-vax@NetBSD.ORG>
From: None <duanne.thompson@tswi.com>
List: port-vax
Date: 10/20/1997 09:41:47
To: brianc@carpediem.com (Brian Chase)
From: Ignatios Souvatzis <ignatios@cs.uni-bonn.de>
Reply-to: port-vax@NetBSD.ORG
Cc: port-vax@NetBSD.ORG

Brian Chase wrote:
> 
> I've used Ultra-wide SCSI drives on SGI's which only support fast narrow
> SCSI :-)  Not that that's entirely relevant to original question, but I've
> found that newer SCSI drives tend to be fairly tolerant of older SCSI
> controllers.  I was surprised by this, but maybe backwards compatiblity is
> specified in the newer standards?  I guess I should browse the SCSI FAQ.

Yes and yes (its specified in the standard, and you should browse the
SCSI FAQ :-).

For synchronous transfer speed (actually cycle time), synchronous transfer
fifo depth (Network Technicians would call this window size) and bus width,
an Initiator and each of its Targets negotiate (at least on) the initial 
contact after a SCSI bus reset.

- if the other machine does reject the negotiation MSG, don't do it,
  that is do async transfers or byte-wide transfers, respectively.
- if the other machine wants bigger cycle time, less fifo depth or less
  bus width, obey.

- all the configuration messages are byte-wide anyway.

This way it's guaranteed that you can talk to wide disks on a small bus, or
to small disks on a wide bus, and even mixed in any combination.

Regards,
	Ignatios



--simple boundary--