SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    Re: iSCSI: default values



    Black_David@emc.com wrote:
    > 
    > Default values may be difficult to define for text keys added after
    > the first stable version of the spec.  The problem is that an
    > originator who does not send a text key cannot distinguish between:
    > 
    > - a responder who understands the key and implements the default AND
    > - a responder who does not understand the key and does something
    >         acceptable under a version of the spec prior to introduction
    >         of the key.
    
    David,
    
    I see a minor nit in the above described scenario.
    
    During iscsi login, both sides need to agree on a common overlapping
    version in their respective ranges of (version_low, version_high) sent
    in the login pdu. Lack of a common version results in the target
    rejecting the login with an "unsupported version" status explanation.
    
    Once the 2 sides have agreed to a common version, both sides know all
    the supported operational & security keys in use. Thus, one side cannot
    send a "key defined in a later version of the spec" which the other side
    does not understand.
    
    Such an erroneous key usage, would have to be due to a buggy
    implementation and would cause the other side to treat it as a vendor
    unique key being originated to which it would respond with
    NotUnderstood. 
    
    That said, I agree with the intent of your suggestions below.
    
    Thanks,
    Santosh
    
    
    > This whole notion of omitting a key being equivalent to sending
    > it with its default value strikes me as potentially problematic.
    > I would have thought that omitting a text key was equivalent to
    > "Don't Care" with the default values being "RECOMMENDED" rather
    > than "MANDATORY".  This leads to the following rule, which I think
    > should go in regardless:
    > 
    > - An implementation whose operational correctness and/or performance
    >         depends on the value of a text key MUST explicitly negotiate
    >         that key.
    > 
    > This is somewhat like the old adage about voting - those who do not
    > vote should not complain about the results, and leads to more robust
    > negotiation exchanges because mismatched assumptions about defaults
    > get exposed.
    > 
    > I realize this yields additional messages in some cases, but would
    > suggest that the additional negotiation robustness is an offsetting
    > benefit.
    > 
    > Comments?
    > --David
    > ---------------------------------------------------
    > David L. Black, Senior Technologist
    > EMC Corporation, 42 South St., Hopkinton, MA  01748
    > +1 (508) 435-1000 x75140     FAX: +1 (508) 497-8500
    > black_david@emc.com       Mobile: +1 (978) 394-7754
    > ---------------------------------------------------
    
    -- 
    ##################################
    Santosh Rao
    Software Design Engineer,
    HP-UX iSCSI Driver Team,
    Hewlett Packard, Cupertino.
    email : santoshr@cup.hp.com
    Phone : 408-447-3751
    ##################################
    


Home

Last updated: Wed Oct 17 15:17:27 2001
7269 messages in chronological order