Re: iSCSI Change - Login phase

[Mark Bakke <mbakke@cisco.com>]

Julian and Matthew-

This is looking good; I have just a few comments:

Julian Satran wrote:
> 1.1.2     I - Immediate
> 
>    Login SHOULD be issued as an immediate request (I=1) except for the
>    leading login phase in which all requests that MUST have I=0.

Did you mean "leading connection"?  If so, the only thing that is
special about the leading connection is that it needs to communicate
the initial CmdSN value; there is no reason to use I=0 for the
login requests themselves.

To make this interoperate easier, I would like to get rid of
the SHOULD.  How about:

   Login MUST be issued as an immediate request (I=1) except for the
   leading login request of the leading connection in a session, which
   MUST have I=0 and a valid CmdSN.

I think that this would be simpler, and still get what we need,
which is to communicate a CmdSN to start with when we hit full
feature phase.

> 1.1.8     CmdSN
> 
>    CmdSN is either the initial command sequence number of a session (for
>    the first Login of a session - the "leading" login) or the command
>    sequence number in the command stream (e.g., if the leading login
>    carries the CmdSN 123 the next non-immediate command carries the number
>    124 etc.).

If we make the above change regarding the Immediate bit, this changes to:

   CmdSN is the initial command sequence number of a session, and is
   valid on the first login request of the leading connection in a
   session.  This determines the CmdSN used in the next non-immediate
   command in the session (e.g., if the leading login request carries
   the CmdSN 123 the next non-immediate command carries the number
   124 etc.).



-- 
Mark A. Bakke
Cisco Systems
mbakke@cisco.com
763.398.1054