Re: iSCSI : SCSI Response PDU

[Matt Wakeley <matt_wakeley@agilent.com>]

julian_satran@il.ibm.com wrote:
> 
> responses in text:
> 
> Thanks,
> Julo
> 
> Santosh Rao <santoshr@cup.hp.com> on 10/03/2001 01:11:23
> 
> Please respond to Santosh Rao <santoshr@cup.hp.com>
> 
> To:   ips@ece.cmu.edu (ips)
> cc:   santoshr@hpcuhe.cup.hp.com (Santosh Rao)
> Subject:  iSCSI : SCSI Response PDU
> 
> Julian/All,
> 
> Some comments on recent changes to the SCSI Response PDU in Rev 04/05 :
> 
> 1) Sense Length has been removed. This field is required and should be
> re-introduced. (Can anyone comment on the reasons behind its removal(?) ).
> 
> +++ sense is in the data segment - data segment length is sense length +++

I figured that out, but there is nothing in the document stating that.  Since
this is a public spec, it should spell everything out (otherwise you, as the
author, will constantly get questions).

> 2) There is a reference to Response Data [without a corresponding
> response length. Once again, response length is required, if a
> response data is used.]. Response Data is modelled on the FC
> response data and is useless in the current context.
> (The FC response data only contained a response code).
> Any target that wishes to send vendor unique information
> could do so through vendor unique sense data.
> 
> All references to Response Data should be removed.
> +++ Responses are not sense +++
> 3) A range of iSCSI Response codes should be assigned for vendor-unique
> responses.
> +++ Good Idea +++

Yeah, will contribute to interoperability... NOT!

> 4) The spec should state explicitly whether implementors are to use the
> REJECT PDU or a Scsi Response PDU with iSCSI Response code to indicate
> failure to process a command due to invalid command pdu fields. If the
> SCSI Response PDU is to be used to convey this, an additional iSCSI
> response code must be added for "Invalid Command PDU".
> +++ section 6.1 covers this - Matt Wakeley proposed changing the name to
> protocol errors +++
> 5) In section 2.5.1, pg 40, the spec states :
> "For all these functions, the SCSI Task Management Response MUST be
> returned using the Initiator Task Tag to identify the operation for
> which it is responding."
> 
> The above MUST should be changed to SHOULD since a task mgmt command can be
> responded to with a REJECT response on a format error.
> +++ I will make " for all these functions, if executed ... +++