Re: iscsi : numerical negotiation wording is ambiguous

[Santosh Rao <santoshr@cup.hp.com>]

Hi Julian,

I think a number of developers [like Matthew B stated it] would be
thinking it is the alternate method that is to be used. The reason being
that mass storage negotiation [FC PLOGI, FC PRLI, SCSI Mode Select, etc]
expect the responder to return the result of the negotiation and
the initiator only parses the response to determine if the result of the
negotiation is acceptable to it.

The above approach is a familiar one for most mass storage folks,
involves 
less code [only 1 side doing the determination of the result], is a
simpler method, allows less room for error and eases debugging.

I request that the above proposal be considered and the wording
of the negotiation tightened to allow no room for any
mis-interpretation.

Thanks,
Santosh


Julian Satran wrote:
> 
> Santosh,
> 
> I understood what you wording means but I am not sure that we want all the
> side-effects.
> The negotiation as defined  now allows both parties requester or responder
> to state their wishes and the LAW
> insatiate the result in both.
> 
> Your wording means that the responder selects the value according to the
> rule. What if the responder is either a rogue or
> just a simple minded target.  Let me give an example:
> 
> I am building a simple minded target that has an 8K buffer and says  always
> (has it wired) DataPDULength=8192
> in its first Login response (that is his buffer).
> 
> If an initiator sends him as a "offer" or as a "responder" 16192 then with
> the current wording things are fine and both will
> have settled to 8192.
> 
> If the initiator sends an offer of 4096 and the target gives his (only
> thing he knows) 8192 it is still fine - both select 4096.
> 
> With your wording some of the negotiations will fail since you assume that
> the rule should be expressed in building the answer and not in selecting
> the result.
> 
> In the end in both case you have to do selections at both target and
> initiator but the current rule enables simple-minded prewired messages
> while your does not (the responder message defines the selection and the
> offerer has to check it).
> 
> Sorry for this long message for such a simple question.
> 
> Julo
> 
> 
>                     Santosh Rao
>                     <santoshr@cup.       To:     ips@ece.cmu.edu
>                     hp.com>              cc:
>                     Sent by:             Subject:     Re: iscsi : numerical negotiation wording
>                     owner-ips@ece.        is ambiguous
>                     cmu.edu
> 
> 
>                     26-09-01 23:16
>                     Please respond
>                     to Santosh Rao
> 
> 
> 
> Julian,
> 
> What is the responding party supposed to offer ? Is it supposed to
> determine the result of the
> negotiation (higher or lower value, as the case may be) and send that as
> its response ?
> 
> Or, is it supposed to send in its numerical value and the initiator picks
> the higher or lower of
> the 2 ?
> 
> This does'nt come across clear enough in the definition and is open to
> mis-interpretation. Please
> see the suggested re-word in its place.
> 
> Thanks,
> Santosh
> 
> Julian Satran wrote:
> 
> > Santosh,
> >
> > I am missing something. The rule states what value both parties should
> have
> > after both have seen the two values.
> > Obviously we assume that no error occurs and the responder value is seen
> y
> > the offering party or the negotiation fails.
> >
> > What exactly is ambiguous about it?
> >
> > Julo
> >
> >
> >                     Santosh Rao
> >                     <santoshr@cup.       To:     ips@ece.cmu.edu (ips)
> >                     hp.com>              cc:
> >                     Sent by:             Subject:     iscsi : numerical
> negotiation wording is
> >                     owner-ips@ece.        ambiguous
> >                     cmu.edu
> >
> >
> >                     26-09-01 19:59
> >                     Please respond
> >                     to Santosh Rao
> >
> >
> >
> > Julian & All,
> >
> > The definition of numerical negotiation in Section 2.2.4 of Rev 7.97
> > reads :
> >
> > "In numerical negotiations, the offering and responding party state
> >  a numerical value. The result of the negotiation is key dependent;
> >  frequently the lower or the higher of the two values is used."
> >
> > The above definition is ambiguous, since it does not specify whether it
> is
> > the originator or the responder that computes the result of the
> > negotiation.
> >
> > i.e. Is it the responsibility of the target to pick the higher or lower
> of
> > the 2 values and respond with the result of the negotiation ?
> >
> > OR :
> > Is it the originator that has to pick the result of the negotiation
> > based on the key it sent and the key it got back ?
> >
> > I would suggest that the wording be clarified to indicate that the
> > responder picks the result of the negotiation and sends this result back
> > in its response for this key.
> >
> > Perhaps, some re-wording along the following lines may be in order :
> >
> > "In numerical negotiations, the offering party states a numerical
> >  value, and the responding party states the result (operational value)
> >  after the negotiation.  The result of the negotiation is key
> >  dependent; responder determines it based on the lower or the higher
> >  of the two values - offering party's value, and what the responder
> >  can support."
> >
> > Comments ?
> >
> > Regards,
> > Santosh
> >
> > --
> > #################################
> > Santosh Rao
> > Software Design Engineer,
> > HP, Cupertino.
> > email : santoshr@cup.hp.com
> > Phone : 408-447-3751
> > #################################
> 
> #### santoshr.vcf has been removed from this note on September 27 2001 by
> Julian Satran

-- 
##################################
Santosh Rao
Software Design Engineer,
HP-UX iSCSI Driver Team,
Hewlett Packard, Cupertino.
email : santoshr@cup.hp.com
Phone : 408-447-3751
##################################