Re: [Roll] RPL clarification needed: moving down the DAG

["Pascal Thubert (pthubert)" <pthubert at cisco.com>]

Hi Mukul:

We currently have a rule NOT TO follow a parent that falls down a current DODAG iteration. This was added to avoid count to infinity very early in RPL. A node still can and should IMHO follow a parent that jumps if that node looses all parents in its current iteration.

Pascal

>-----Original Message-----
>From: Mukul Goyal [mailto:mukul at uwm.edu]
>Sent: jeudi 5 novembre 2009 13:05
>To: Richard Kelsey
>Cc: Pascal Thubert (pthubert); roll at ietf.org
>Subject: Re: [Roll] RPL clarification needed: moving down the DAG
>
>Hi Richard
>
>If the node is in a "count to infinity" situation in DAG1 and is a part of DAG2 as well (or can join DAG2
>quickly if it wants to):
>
>1) it will start using DAG2 for packet forwarding once DAG2 becomes more attractive than DAG1.
>
>2) But how does the "count to infinity" the node is doing in DAG1 affect DAG2 otherwise? I can see that
>frequent generation of DAG1 DIOs by the in-loop DAG1 nodes will cause extra contention for the radio channel,
>but what else may happen?
>
>Thanks
>Mukul
>
>----- Original Message -----
>From: "Richard Kelsey" <richard.kelsey at ember.com>
>To: "Mukul Goyal" <mukul at uwm.edu>
>Cc: pthubert at cisco.com, roll at ietf.org
>Sent: Thursday, November 5, 2009 5:31:16 AM GMT -06:00 US/Canada Central
>Subject: Re: [Roll] RPL clarification needed: moving down the DAG
>
>   Date: Wed, 4 Nov 2009 17:32:05 -0600 (CST)
>   From: Mukul Goyal <mukul at uwm.edu>
>
>   2. Now, lets suppose that new parent (B) has a higher rank than the
>   node's current rank and is infact in the node's sub-DAG. The node (A)
>   chooses it as a parent and consequently increases its rank. Now, we
>   have a routing loop. [...] We will have "count to infinity" situation only
>   when nodes in the loop are isolated from rest of the network. In that
>   case, they will do "count to infinity" and produce lot of DIOs but we
>   should not really care about these nodes (and their DIOs) since they
>   are isolated any way.
>
>Hi Mukul,
>
>Remember that there may be multiple DAG instances.  Being
>isolated from one DAG root does not mean that a node has
>nothing useful to do.  It may just mean that it will use
>a different DAG for the moment.
>                                     -Richard Kelsey