Robert Frey wrote:
>
> Dan Jones wrote:
>
> > Robert Frey wrote:
> > >
> > > I agree it's better to have a system running than not running, but I would also
> > > argue that domain validation speed fall-back is a symptom of something wrong
> > > with the SCSI "domain" that should be fixed.
> > >
> > Domain validation covers two cases. You are referring to a system
> > that does not work as specified. The other case for domain
> > validation is to support legacy equipment, which still has
> > economic value, but cannot run reliably at the fastest SCSI rates.
>
> Every device negotiates its data transfer speed independently with the host using
> WDTR, SDTR, and PPR messages. Legacy devices never even try to run faster than their
> maximum speed that is fixed at the time the device is manufactured. Legacy devices
> aren't affected by other devices running at faster speeds, because at any time the
> bus is only controlled by a single target and initiator. All other targets just stay
> off the bus and don't care about what data rate is used by the target and initiator
> that happen to currently control the bus.
>
True, but beside the point. You might check back and notice that I
did not refer to SCSI devices. SCSI backplanes are designed and, more
importantly, tested to the performance levels that are available to
the designer. Cables could be included in this. Any system that was
released over a year ago cannot be guaranteed to run at U160 even if
you should upgrade drives and controller. The devices and controller
would negotiate for a transfer rate that could not be successfully
used.
> Also a SCSI bus affected by external factors like signal interference is a system I
> would argue is not correctly specified. If there is signal interference why not
> remove it? Maybe I'm missing something but I think systems can be specified and
> designed taking into account these factors.
>
I, personally, wouldn't like to use a signal interference arguement
for domain validation.
> I like in theory the idea of domain validation. I'm just worried about the
> implementation and application. It shouldn't be used to cover up design flaws or
> misconfiguration. Instead it should be used as a way to notify the end
> user/administrator of a problem and allow continued sub-optimal operation. If that's
> the implementation I think it's very useful.
>
> Linux because of its flexibility clearly has an opportunity to do what I think is a
> good implementation. As opposed to Windows NT which I think will have a lot of
> difficulty ever doing a good implementation. NT HBA drivers have no way other than
> creating event log messages to provide an administrator information.
>
> Bob Frey
> [EMAIL PROTECTED]
--
Dan Jones, Storage Engineer VA Linux Systems
V:(408)542-5737 F:(408)745-9130 1382 Bordeaux Drive
[EMAIL PROTECTED] Sunnyvale, CA 94089
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
