SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    RE: iSCSI: Login Request error



    --- Bill Studenmund <wrstuden@wasabisystems.com>
    wrote:
    
    > I'm not sure about renegotiating REJECTED keys. I
    > think it'd depend on
    > what the likely reasons for rejection are, and how
    > likely they will be
    > different the second time.
    
    Aha! :-) (Pat has actually already described almost
    word by word what I was planning to say about any
    reasons given, but I'll do the same anyway...)
    So we have to look at what the reasons could be.
    
    Junk value on binary key? Why was it sent in the
    first place? Inadmissible numerical value, string
    literal, or range sent? Again, why should the
    responder cut any slack to such initiators and let
    them try again? A list sent consisting of values 
    that are not agreeable to the responder? If there
    are more values that the originator could offer,
    why weren't they included in the first request?
    The only understandable reason I can imagine is
    for trying out numeric ranges, one after another
    in the order of preference. But if this is the
    reason for allowing renegotiations, then we'd be
    better off introducing lists of ranges, a thing that
    has been discussed here, but was decided as yet
    unnecessary. So this could not be the reason.
    
    There is a somewhat risky use for sending
    a junk value however---it can simply be used to let
    the other side pick something w/o restrictions. But
    there is no guarantee that the other side will not
    Reject it and consider the whole negotiation
    sequence failed with that (which in my understanding
    (and I could dig up the relevant messages on the
    list) is also a valid behavior).
    
    > I guess irrelevant makes sense for things like
    > negotiating marker spacing
    > when you don't support markers.
    
    But we don't have any dependencies between parameters
    anymore, and that includes the marker-related keys.
    It is quite possible to view them as independent.
    It is alright to negotiate the marker interval but
    not turn on the use of markers.
    (By using 0 as a special interval value, we 
    could also eliminate the boolean marker keys,
    but that's unimportant).
    
    Anyway, thanks for keeping the discussion up!
    
      Martins Krikis, Intel Corp.
    
    Disclaimer: these opinions are mine and may not be
                those of my employer
    
    
    
    __________________________________________________
    Do You Yahoo!?
    LAUNCH - Your Yahoo! Music Experience
    http://launch.yahoo.com
    


Home

Last updated: Wed May 22 21:18:27 2002
10227 messages in chronological order