Re: iSCSI 4.1 & 4.2

[Luben Tuikov <luben@splentec.com>]

pat_thaler@agilent.com wrote:
> 
> I agree. Negotiations are not stateless. At a minimum there is the state of
> the latest value plus knowledge of whether that value has been negotiated
> yet.

You mean:
_offers_ MUST have a state as well as
responses MUST have a state (at which point we have
a complete negotiation sequence which inherits
the state of the offer).

http://www.pdl.cmu.edu/mailinglists/ips/mail/msg09751.html

Mallikarjun, this is a stronger rule than the
one you posted, but due to its more
restricting nature, makes negotiation outcomes
more predictable.

This weeds out some incosistencies and loops,
given that the party who sent a ``key=Reject''
is not expected to take any further
actions regarding that ``key'',
but may close the connection as it sees fit.

http://www.pdl.cmu.edu/mailinglists/ips/mail/msg09771.html

I think Julian might've already added this.

(Its ok for something to be incomplete,
but when incosistent, it is useless. -- Me)

-- 
Luben