SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    Re: multiple connections



    David,
    
    Black_David@emc.com wrote:
    
    > Julian,
    >
    > I have to ask you not to do that.  There is consensus
    > in the WG that multiple connection sessions are an
    > important feature that needs to be specified, but
    > should be optional to implement.
    
    agreed.
    
    > There is NOT consensus
    > on what the right design approach is.
    
    agreed.
    
    > I have off-line
    > email from proponents of both the Asymmetric and Symmetric
    > multi-connection session models expressing dismay
    > at the separation of them from the main specification
    
    Exactly.  There has been NO consensus that multi-connection sessions should be
    removed from the main specification (only your proposal to do that).  Just
    because we can't (yet) agree on something, does not mean "let's not do it at
    all, or put it off until later."
    
    Any well performing iSCSI implementation will have to have some level of
    hardware implementation.  Putting off the discussion of multiple connections
    will either delay hardware, or require hardware rolls to implement the new
    functionality after the initial hardware is out.
    
    >
    > and arguing that their preferred approach is the right
    > one.  These reinforce my observations that there is
    > no consensus on the issue, and that the mailing list
    > discussion is unlikely to achieve consensus.  As I
    > stated in earlier email, the requirement for
    > multiple connection sessions has not been removed,
    > but spending the next 6-8 weeks discussing it on the
    > mailing list does not appear likely to achieve consensus.
    > We need to try something else, namely an off-line
    > design team.
    
    We already had the offline design team and we came up with something that
    everyone is arguing about.  So let's have some more discussion on what it is
    that people don't like, and what the requirements are, so that another offline
    design team has something to work with.
    
    > In the near term, list bandwidth is
    > better used to make progress on issues where
    > progress is still possible, such as flow control.
    
    In my mind, the flow control is directly related to how many connections there
    are, and how they will work.
    
    >
    >
    > Thanks,
    > --David
    
    Matt Wakeley
    Agilent Technologies
    
    


Home

Last updated: Tue Sep 04 01:07:06 2001
6315 messages in chronological order