Hi Roland,

Thanks for the clarification. 

Acee 

On 11/12/22, 10:43 AM, "Bless, Roland (TM)" <[email protected]> wrote:

    Hi Acee,

    I just reviewed the video recording of the RTGWG session,
    because my colleague told me that we probably have talked past each
    other... :-) Here is a small clarification on your question.

    So your scenario was that the node X (see figure below)
    sending out the packet to the yet unknown destination node Z does
    not know a priori whether the destination Z is in the left or right
    partition.
        _________           ________
       /         \         /        \
      |     Z     |--(X)--|    Y     |
      |           |       |          |
       \_________/         \________/

    My answer was that the ID-wise closest contact (to the destination node
    ID Z) determines which interface (let's assume for node X it would be Y)
    is used and that is correct.
    But that may also mean that the packet first travels into the "wrong"
    direction. This is the stretch for the first packets that we have to
    accept as tradeoff for our scalability. Later packets would not be
    routed into the wrong direction any more, because cycles would have
    been removed from the answer and the later path.

    So my answer that the NodeID tells you into which direction X has to
    route is in that sense correct that it may also be the "wrong"
    direction. Our NodeIDs have indeed no encoded dependency on the 
    topology, so node X cannot infer from the ID Z in which partition it
    is in.

    Hope you have/had a safe trip home.

    Best regards,
      Roland


_______________________________________________
rtgwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rtgwg

Reply via email to