[Roll] MPL leaf node behaviour

[James Ko <[e-mail]>]

During testing of my implementation of MPL I found that the leaf nodes in a mesh network (or the last MPL Forwarder to receive a multi-cast packet) would continue to retransmit at every trickle time 't', until expiration 'e' is reached, since it does not receive the same number of 'consistent' transmissions from a neighbour and count 'c' never reaches the consistency constant 'k'.


Example with k = 2, seq [1], nodes <n>

Node <1>
        Node <2>
        Node <3>

TX[1], c = 0
        RX[1] from <1>, c = 0


RX[1] from <2>, c=1
        TX[1], c = 0
        RX[1] from <2>, c = 0

TX[1],  c = 1
        RX[1], from <1>, c = 1
        TX[1], c = 0

RX[1] from <2>, c = 2
        TX[1], c = 1
        RX[1] from <2>, c = 1



RX[1] from <3>, c = 2
        TX[1], c = 1


        RX[1] from <3>,  c = 3
        TX[1], c = 1



RX[1] from <3>,  c = 4
        TX[1], c = 1







Node <3> would continue to retransmit [1] until 'e' expirations.  Even if reactive forwarding is enabled (CONTROL_MSG_EXPIRATIONS != 0) and a 'consistent' control message is received it does not increment 'c' for the data messages.  The data message trickle is reset only for 'inconsistent' messages identified by the control message.


If the expirations 'e' is large, what, if any, strategies could be used to prevent the leaf node from continuing to transmit since it never receive k number of consistent transmissions?


One possible strategy is if reactive forwarding is also enabled then a data message should only be transmitted if 'c' < 'k' for both the data message and the control message.


Of course, this leaf node consistency problem also exists for the control messages following the trickle algorithm.


James