[Unstrung-hackers] RPL requirements

Vijay Anand R vjanandrgky at gmail.com
Thu Oct 15 06:26:23 EDT 2015


Hi Michael,

I believe we have two IDs to uniquely identify a RPL DODAG.

a. The instance ID - We could have a common Global instance ID like what 
you suggested for all AN domains. But this is only 7+1 bits, so maximum 
of 128 values.
b. The DODAGID - according to RFC6550, this seems to be the root 
identifier, and in my humble opinion the domain specific ULA address of 
the root would be a good identifier within the AN domain DODAG.

On 10/14/2015 8:52 AM, Michael Richardson wrote:
> Vijay Anand R <vjanandrgky at gmail.com> wrote:
>> Design Considerations
>> ====================
>>
>> a. The DODAG root is administratively configured to be the
>> controller/registrar host.
>> b. The DIO message used to build the DODAG would be configured as below.
>>      i) The 128 bit DODAGID is the IPv6 Domain ULA address of the host device.
> Is it your goal to automatically configure this?
> I think that is a mistake: other hosts then have to figure out what DODAGID
> to listen to.  I'm not really sure why in the end we have DODAGID looking
> like an IPv6 address, it's just 16 bytes.  Instead, I suggest that it be
>       414e 494d 4141 4e49 4d41 414e 494d 4141
>
> (Which is just ANIMA in ASCII repeated)
>
>>      ii) PIO option is not required to be distributed to the forwarding
>> elements. The ULA address assignment is handled by AN control plane protocol.
> The PIO could be turned off, sure.
>
>>      iii) The DODAG operates in Storing Mode of Operation with multicast
>> support. P2P traffic is not required to be routed all the way to the DODAG
>> root.
> sure.
>
>> c. AN is more concerned with the connectivity of devices and hence not
>> particular about a objective function. OF0 as defined in RF 6552 is suitable
>> enough. Therefore the root and hence the DODAG is grounded.
> yes.
>
>> d. The preference bit is the same for all nodes (could be 3).
>> e. The Domain ULA address is advertised in the DAO messages as reachability
>> information.
> That's done with the tree at:
>         https://github.com/mcr/unstrung/tree/t10143-prefixes-from-interfaces
>
> It's failing in Travis-CI, because it seems that the 12.04 libboost-dev
> does not have binomial-heaps available. I'm deciding what to do about that.
>
>> f. The DAO messages are exchanged over the Overlay network over the IPSEC/GRE
>> tunnels. Therefore the DODAG is constructed over overlay network.
> Ideally, we need to have -i take a wildcard, and learn operate automatically
> on new interfaces.
>
>> g. Any new tunnel interface that is created by AN should be associated with
>> RPL.
>> h. Election of a new ROOT when the current ROOT goes down could be based on
>> the lowest prefixed IPV6 Domain ULA address. This is desirable so that the
>> ULA address are assigned incrementally and the node closest to the current
>> ROOT could potentially become the new ROOT. This is an implementation detail
>> and is not covered in any of the RFCs.
> RPL already creates new roots if you use the same DODAGID. There is no need
> for a new protocol to perform an election.
>
> --
> ]               Never tell me the odds!                 | ipv6 mesh networks [
> ]   Michael Richardson, Sandelman Software Works        | network architect  [
> ]     mcr at sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [
>



More information about the Unstrung-hackers mailing list