Re: [Anima] ACP RPL root

[Benjamin Kaduk <[e-mail]>]

On Tue, Jul 07, 2020 at 09:47:59PM -0400, Michael Richardson wrote:
> 
> Ben, is this DISCUSS comment still alive?

>From memory, I think it's been resolved.

I'm sure that a discuss about whether/how the RPL root is automatically
determined got resolved for *some* document, but am not 100% sure if it was
this one or one of the 6TiSCH ones.  In any case, the discussions that
preceded that resolution also served to educate me about some of the finer
points that I had completely missed with my original comment, and that
education (combined with your description below) convinced me that there's
not a (certainly discuss-level, and probably not any) issue here.

-Ben

> > Section 6.11.1.14 places a normative ("SHOULD") requirement on the RPL
> > root, but if I understand correctly the RPL root is automatically
> > determined within the ACP, and thus the operator does not a priori know
> > which node will become the RPL root.  Am I misunderstanding, or is this
> > effectively placing this requirement on all ACP nodes?
> 
> In LLNs, it is common to designate a specific, high-capacity node as the RPL
> root.
> 
> In particular, in non-storing mode, the RPL root has to keep a RIB and FIB
> for every single node, while the other nodes need only three to five FIB
> entries to keep track of parents and active children.
> 
> We are using storing mode, so every node already has to have the capacity to
> have a RIB and FIB on the order of the number of ACP nodes.
> This is not arduous for an Enterprise or ISP class router.
> That's the major cost of being the root: the root has to have a complete
> table. (But, in storing mode, we only have next-hop, not the full path)
> 
> As originally envisioned, the automonic network should be self-forming and
> self-healing, which means that partitions of the network should recognize
> that there is no root, and choose one.  That level of autonomy is not
> reached in this document, or in the ANIMA architecture. In fact, that
> level of autonomy got spun off as the SUPA WG.
> 
> So, yes, in theory, all ACP nodes could be the RPL root, and when they do,
> they need to install blackhole route for ULA space fd00::/10 in order to
> prevent loops.  This shouldn't be a big deal.
> 
> The RPL DODOG root will in mature situations be the ACP registrar,
> and while the market is less mature, in an router designated as the
> ACP-connect.
> 
> --
> Michael Richardson <[e-mail]>, Sandelman Software Works
>  -= IPv6 IoT consulting =-