SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    Re: iSCSI 4.1 & 4.2



    --- Luben Tuikov <luben@splentec.com> wrote:
    > I don't really object to the implicit result of some
    > booleans.
    
    I do.
    
    > 1) Couldn't find incosistencies with other things
    >    I/we mentioned (states of offers, states of
    > responses = states
    >    of complete neg. seq.). That is, couldn't find a
    >    (counter-) example.
    
    My way of figuring out whether a value has been
    negotiated is to see whether it has been sent,
    whether it has been received, and of course I do
    consistency checks. The ability to omit responses
    for boolean values under certain conditions makes
    this task more complicated. I also have to consider
    the value as negotiated when it has been sent, has
    not been received, but the function is OR and the 
    value is Yes, or when the value is AND and the value
    is No. Yes, it can be done, and it can be done also by
    using more variables for values or introducing more
    states, and probably by lots of other means. The
    point is that this does introduce a complication,
    at least in my implementation. And it certainly
    does not help the clarity of the specification.
    
    > 2) For the same reason you all considered that some
    > boolean
    >    responses to be optional: one can determine the
    > outcome of the
    >    negotiation sequence and the state of the whole
    > thing
    >    if one knows f(a,b) and b for certain f()'s over
    > booleans.
    
    Yes, I can determine the value. Just because I can
    doesn't mean that boolean negotiations with omissions
    are simpler than truly explicit boolean negotiations
    (like all the others).
    
    > Keeping the optional responses greatly simplifies
    > things.
    
    No it does not. Complicates the description and
    complicates at least some (otherwise reasonable)
    implementations.
    
    > That is, the sender can safely assume a state and
    > result
    > of the negotiation sequence and SHOULD a reply to
    > that
    > key=value arrive and its value is NOT what was
    > assumed,
    > then the Target should close the connection with
    > Initiator Error
    > and the Initiator should just drop the connection.
    
    Yes, but the same is true for explicit negotiations.
    If an incorrect value is received, I can drop the
    connection. But at least I know that it MUST be
    eventually received. Less MAYs, less choice, less
    room for misunderstandings and error. Simpler
    specification as well. Let's not forget that it
    still (so far incorrectly) claims that negotiations
    are explicit and that the description of which value
    is chosen on omissions is (in my view and nobody
    has disputed it yet) incorrect.
    
      Martins Krikis, Intel Corp.
    
    Disclaimer: these opinions are mine and may not be
                those of my employer.
    
    
    __________________________________________________
    Do You Yahoo!?
    Yahoo! Health - your guide to health and wellness
    http://health.yahoo.com
    


Home

Last updated: Tue Apr 30 14:18:29 2002
9889 messages in chronological order