RE: iSCSI: Why is ACA optional?

["Douglas Otis" <dotis@sanlight.net>]

Charles,

The concept is to prevent a need to filter CDBs going to devices if the
application and the device do not both support ACA.  To meet the needs of
the iSCSI Auto-Sense requirements while not actually obtaining sense
information from the device would be a technique that would allow a means to
satisfy both requirements without upsetting a CA device.  Status should be
enough in this case where Sense would then be obtained explicitly.  SAM
allows this.  The device would make the decision not to include Sense.  The
adapter should know if CA and not ACA is supported and make the choice to
either fake ACA and thus require CDB fiddling or just not send Sense within
the PDU.  In the latter case, simply allowing that behavior provides the
option at the device as how best to handle the situation.  The device then
is not mandated to support ACA and yet still operates correctly in the case
of an error according to an AutoSense requirement (well almost).  These
older legacy devices do not adopt newer standards at the same rate as disks
together with older applications that have been creating dusty libraries.
Should this be a problem with someone, then insist on always fiddling CDBs
and faking ACA.

We have run around in circles on this subject in the past and so I am
nervous about even breaching the subject once again.

> Hi Doug:
>
> > -----Original Message-----
> > From: Douglas Otis [mailto:dotis@sanlight.net]
> > Sent: Thursday, November 30, 2000 2:01 PM
> > To: Charles Monia; julian_satran@il.ibm.com; ips@ece.cmu.edu
> > Subject: RE: iSCSI: Why is ACA optional?
> >
> >
> > Charles,
> >
> > For tape applications, not providing Sense Information within
> > the iSCSI
> > response should remove much of the troubles caused by an ACA
> > mandate.  In
> > otherwords, not seeing Sense within the Auto-Sense PDU should
> > not be seen as
> > an error and thus all is well.
> >
>
> If anything, providing autosense information with the CHECK
> CONDITION status
> mitigates (but does not eliminate) the need for ACA.  With regard
> to tapes,
> ACA is probably going to be essential when such devices get around to
> supporting command queuing.
>
> Charles
>
>
>