Re: [iSCSI]: Key negotiation procedure proposal

[Luben Tuikov <luben@splentec.com>]

Paul Koning wrote:
> 
> I think it's unnecessarily complex.  The "simple implementation" alone
> is plenty.

Certainly!

But if we want more than that, then mathematics tells me
that those are the rules. (I.e. those are the minimum
consistent set of rules which achieve the task at hand.
It also tells me that there is no ``in the middle''
implementation.)

Please also note that ``simple implementations''
can co-exist with ``complete impementations''.
It's just that complete ones will get the TCP FIN,
when rejecting a simple implementation's key=value
pair, or just TCP FIN if the simple implementation
rejected theirs.

Julian, most of the negotiation the way we have it
now, already observes most of those rules.

And as to your request ``if they can be proven
useful'' -- a few examples on your own will
convince even the most doubtful heart. :-)

-- 
Luben
P.S. I just don't want to end up writing a paper
about this (co-authors welcome) as it happened
for CRC32C.