Re: [6lo] Adaption of ROLL for mesh-under

In my case, it's because the network I'm starting with already operates well in a mesh-under regime. Small changes to fundamental mechanisms can have very interesting consequences IRW, so adjusting fundamentals to RFC-defined messages while keeping the same principle of operation would be a relatively low-touch way to move toward the benefits of standard vernacular, use of extensions, etc.  There is a marginal benefit of having possibly slightly smaller messages due to 64-bit IIDs instead of 128-bit IPv6 addresses, which on the whole is not important but out on the fringe of the network every bit counts.  Additionally, SSNI isn't the only company with large mesh-under deployments, so I do believe it could be a shared interest and worthy of taking up.

Benjamin Damm
O: 669-770-4000
E: bdamm@ssni.com  www.ssni.com

________________________________________
From: Carsten Bormann <cabo@tzi.org>
Sent: Tuesday, April 11, 2017 9:46 PM
To: Benjamin Damm
Cc: lo
Subject: Re: [6lo] Adaption of ROLL for mesh-under

Hi Benjamin,

Just speculating here:

In RPL itself, you could use "IP addresses" made up from MAC addresses as IIDs.
For the RPL-routed traffic, you could use 6LoRH-style encapsulation (RFC 8138), which also would fit a mesh-under approach very well.
So I think the total amount of messages that have to be designed/redesigned to make RPL applicable to mesh-under is very low; you’d mainly need a few new code points to make sure router-over RPL/traffic and mesh-under RPL/traffic are not confused.

But this also raises the question why you want to go mesh-under, if in the end everything looks so similar to route-over.

Grüße, Carsten

> On Apr 12, 2017, at 06:33, Benjamin Damm <bdamm@ssni.com> wrote:
>
> Hi folks, I'm new here, so please steer me straight if this is not the right forum.
>
> Is there any work under way that looks at how to use the ROLL-RPL data structures, algorithm, and messages, in a mesh-under network?  The routing in a mesh-under network can be very similar to ROLL-RPL but just one layer down.  Using similar but slightly smaller messages, adjusted for link-level identifiers instead of IP-level identifiers (and of course not supporting IPv6 features in the routing layer) seems like it would be possible, so I'm searching to anyone or any document that might have already led the way for this kind of mapping.
>
> Regards,
> -Ben
>
>
> Benjamin Damm
> O: 669-770-4000
> E: bdamm@ssni.com  www.ssni.com
>
> _______________________________________________
> 6lo mailing list
> 6lo@ietf.org
> https://www.ietf.org/mailman/listinfo/6lo