RE: iSCSI: Negotiation clarifications still needed

[Martins Krikis <mkrikis@yahoo.com>]

--- pat_thaler@agilent.com wrote:

> You don't have to send a declaration but there is no
> reason to 
> prohibit you from doing so.

I know that I don't have to. That's why I said
"should" and not "MUST". I should have said "ought",
perhaps. But there is a reason to prohibit this.
The reason is that then all keys can be treated the
same way regarding whether they can be originated
(or spanned, or whatever, depending on which variation
on this theme we end up with).

> If you would rather not
> check type,
> then don't send any new keys when one is incomplete.

I know. But w/o checking type I can't "nail" the 
other side for breaking the rules :-).

> I suggest that keys that can be sent during
> SecurityNegatiation stage
> should have SN added to Use because use has the
> other information 
> about when a key can be sent.
> 
> I think clearly identifying which keys are not
> subject to negotiation
> is on topic for clarifying negotiation. 

This is all very good, but we didn't have to mix
this issue with the spanning issue. It is IMHO
off-topic w.r.t. the spanning issue, which this
thread has migrated into. For clarifying negotiation,
yes, very much on topic.

> Declarations are involved in the
> spanning/non-spanning issue.

Not unless you start involving them.

> When 
> one has gotten only a partial key one doesn't know
> whether it is 
> a declaration or a negotiation.

OK. And I prefer ignoring whether a key is a
declaration or a negotiation even when I have
received it complete or am planning to originate
it. I have flags telling which sides may use
the key, and I have functions for choosing
a value. I don't have to care about declarations
currently and would love to keep it that way.

I think you're just making your own scheme
subtly more difficult to describe and slightly
more difficult to implement for those that want
to take full advantage of the little additional 
performance benefit that it offers. I see, of
course, that it works. And I can implement it,
and it probably won't even take too long to do
that. The point is that this extra optimization
is yet another (very slight) complication in the
protocol. You are further optimizing a "corner
case", when we really should have opted for the
absolute simplicity in this "corner case".

Martins Krikis, Intel Corp.

Disclaimer: these opinions are mine and may not be
            those of my employer



__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com