Re: iSCSI - Change - Login/Text commands with the binary stage code

["Julian Satran" <Julian_Satran@il.ibm.com>]

Sandeep,

Comments in text.  Julo

Sandeep Joshi <sandeepj@research.bell-labs.com>@ece.cmu.edu on 30-08-2001
19:11:53

Please respond to Sandeep Joshi <sandeepj@research.bell-labs.com>

Sent by:  owner-ips@ece.cmu.edu


To:   ips@ece.cmu.edu
cc:
Subject:  Re: iSCSI - Change - Login/Text commands with the binary stage
      code




Julian,

> > As for the names - I though that security people might object having
the
> > name in clear if the security phase does not make use of the name.
> > Otherwise we can mandate them on the login but I wonder if that is a
real
> > improvement or we are getting carelles.
> >
> > Julo

The problem with these names (and hence the request from Steve and
others
earlier) is that it is not possible to know when the target wants them.

Consider the following excerpts from your latest login proposal..
+++ The initiator know if the authentication method selected needs the name
and will offer it if so (that is mostly context information not related to
the wire protocol +++
> A target MAY use the iSCSI Initiator Name as part of its access control
> mechanism; therefore, the iSCSI Initiator Name MUST be sent before the
> target is required to disclose its LUs.

The above is _very_ confusing..how can the initiator know if the
 the target is doing access control ?
+++ It really is not. LU names can be obtaine only in FFP. The statement
says that you have to disclose them before reaching FFP and explains why
+++
> If the iSCSI Target Name and/or iSCSI Initiator Name is going to be used
> in determining the security mode or it is implicit part of
> authentication, then the iSCSI Target Name and/or iSCSI Initiator Name
> MUST be sent in the login command for the first connection of a session
> to identify the storage endpoint of the session

In both the above cases, how does the initiator know when the target
requires these names?  The partial login response occurs *only* once.
So when going from the security->operational phase, there is no
indication that the target would like these names sent.
+++ the security negotiation is not ended with partial response
but through a complete handshake in the old and new schemes.
The initiator will know before by the selection made if he has to give its
name or not.
The whole point was for us to mandate what is mandatory for iSCSI in
general
and let all the other choices open.
With IPSec it is less of an issue as the name can be presented on an
encrypted although not fully authenticated channel+++

There are 3 options here :
(A) ALways send the names in the login command.  Simplify target
   and initiator and eliminate a few of those partial login
   response codes.
(B) Maintain a configuration database (per-target) of when names
   must be sent - adds an administration burden.
(C) Change the wire protocol to allow the target to indicate when
   the names must be sent - again more complications.
+++ please read and react to the new proposals +++
To round up, I prefer Option (A).  These are just names and not
passwords, so the security risks are minimal.  Are we trying to
protect against traffic analysis ?
+++ I agree that it is simpler to think about it this way. I do not agree
that it is more difficult to implement.
If you need the name in the security negotiation it is just another
parameter of the
method and you will know when to present it.
If you don't need it you will present it on the last Login request (or text
in the old version) - the one that holds the F bit set.
IAgain I hold no strong opinion for or against what you suggest.
I simply don't want us to make a decision based on convenience. It is a bad
guide.
+++
-Sandeep

> >
> > Steve Senum <ssenum@cisco.com>@ece.cmu.edu on 29-08-2001 23:59:36
> >
> > Please respond to Steve Senum <ssenum@cisco.com>
> >
> > Sent by:  owner-ips@ece.cmu.edu
> >
> >
> > To:   ietf-ips <ips@ece.cmu.edu>
> > cc:
> > Subject:  Re: iSCSI - Change - Login/Text commands with the binary
stage
> >       code
> >
> >
> >
> > Julian,
> >
> > A couple of ideas from Matthew Burbridge & Co.'s
> > login proposal that has generated some interest here:
> >
> > 1. Removal of partial login response.  Is it still needed?
> >
> > 2. Requiring Initiator and (if not a discovery session)
> >    Target names on login command, so they are always
> >    available if needed by the initial phase.
> >
> > Comments?
> >
> > Regards,
> > Steve Senum
> >
> >
> >
> >