Re: [Roll] Hosts part of the RPL instance? Re: definition of "RPL Domain"

[Yvonne-Anne Pignolet <[e-mail]>]

I fully agree with Ulrich


Yvonne-Anne Pignolet
Dr. Sc. ETH
ABB Schweiz AG
Segelhofstrasse 1K
5400 Baden-Dättwil, Switzerland
Phone: [phone]
Mobile: [phone]
email: [e-mail]


[e-mail] wrote on 17.11.2011 02:00:06:

> From: Ulrich Herberg <[e-mail]>
> To: Mukul Goyal <[e-mail]>
> Cc: "[e-mail]" <[e-mail]>
> Date: 17.11.2011 01:59
> Subject: Re: [Roll] Hosts part of the RPL instance? Re: definition 
> of "RPL Domain"
> Sent by: [e-mail]
> 
> My 2cts on the various terminology discussions:
> 
> - An "RPL host" seems contradictory to me. Either it is a host, in 
> which case it does not know anything about RPL, or it is an RPL 
> router (leaf node or not, it still remains a router). We should 
> allow for hosts (or be prepared to fight with the IAB for the next 
> years why we think that we should break the IP architecture).
> 
> - A question that comes to my mind: Is it specified anywhere how to 
> add the RPL IP headers for the traffic direction to a data packet 
> from a host received by an RPL router?
> 
> - "RPL domain"; We should just stick to official terminology, i.e, 
> "Routing Domain" in this case. I think it has been specified in RFC1136.
> 
> - RPL traffic: I don't like the term. I would stick to either 
> control traffic or data traffic. Everyone understands these terms. 
> No need to invent new terms.
> 
> Regards
> Ulrich
> 
> On Nov 17, 2011, at 2:38, Mukul Goyal <[e-mail]> wrote:
> 
> > I guess the desired behavior would be:
> > 
> > A host sends out a message to its RPL router. The router adds RPL 
> SRH or RPL option to the IPv6 header and forwards the message 
> further. No need for IP-in-IP tunneling. Any error message comes 
> back to the router and the router handles the message. The host just
> sends and receives messages.
> > 
> > Thanks
> > Mukul
> > 
> > ----- Original Message -----
> > From: "Mukul Goyal" <[e-mail]>
> > To: "Don Sturek" <[e-mail]>
> > Cc: [e-mail]
> > Sent: Wednesday, November 16, 2011 12:29:59 PM
> > Subject: Re: [Roll] definition of "RPL Domain"
> > 
> > Hi Don
> > 
> > I dont want hosts to know about RPL. I just want the RPL routers 
> to consider the hosts as part of the RPL instance so that the RPL 
> router does not have to do IP-in-IP tunneling to forward packets 
> generated by a host.
> > 
> > Thanks
> > Mukul
> > 
> > ----- Original Message -----
> > From: "Don Sturek" <[e-mail]>
> > To: "Mukul Goyal" <[e-mail]>, "Sébastien Dawans" 
> <[e-mail]>
> > Cc: [e-mail]
> > Sent: Wednesday, November 16, 2011 12:22:55 PM
> > Subject: Re: [Roll] definition of "RPL Domain"
> > 
> > Hi Mukul,
> > 
> > I guess my view on this is the opposite of yours.  I would like to see
> > host-only devices not need to know anything about RPL.  Here is why:
> > 1)  Code savings.   Removing RPL from these host only devices would 
allow
> > for deployment on smaller footprint devices
> > 2)  Battery operated devices.   Some host only devices are deployed on
> > non-mains powered devices.  It would be nice for these devices to not 
have
> > to listen for any RPL control messages yet still support transmission 
into
> > a RPL routing domain.
> > 
> > Don
> > 
> > 
> > 
> > On 11/16/11 10:07 AM, "Mukul Goyal" <[e-mail]> wrote:
> > 
> >> Hi Sebastien
> >> 
> >> First, I would like to clarify that the need to define "RPL domain" 
arose
> >> because draft-ietf-6man-rpl-option and 
draft-ietf-6man-rpl-routing-header
> >> were using the term. Now, these drafts use the term "RPL instance" 
and
> >> hence there is no real need to define the term "RPL domain" any more. 
I
> >> will change draft-ietf-roll-p2p-measurement so that all references to
> >> "RPL domain" are changed to "RPL Instance".
> >> 
> >> Now returning to the question whether hosts should be considered part 
of
> >> the RPL Instance, the benefit of doing so is that there is no need to 
use
> >> IP-in-IP tunneling when a host sends out some data. If a host is not
> >> considered part of the RPL Instance, its default RPL router is 
obliged to
> >> use IP-in-IP tunneling to forward the packet further. IP-in-IP 
tunneling
> >> means an extra IPv6 header and thus less space for payload if you 
want to
> >> avoid fragmentation. Also, if the packet is traveling along a DAG, 
the
> >> encapsulation/decapsulation needs to be done at every hop, which 
sounds
> >> fairly heavy duty processing to me.
> >> 
> >> So, I would like to explore if there is a way we could consider hosts 
to
> >> be a part of the RPL Instance.
> >> 
> >> Regards,
> >> Mukul
> >> 
> >>> On what ground would you assume that a non-RPL aware host connected 
to a
> >>> RPL-router (in this case I would call it a border router) is in 
a/the
> >>> RPL Domain?
> >> 
> >>> From what I've seen in the drafts, the term "RPL Domain"'s primary
> >>> purpose it to differentiate the limits of "RPL-aware" nodes for IP
> >>> traffic that needs to transit to or from a set of RPL-aware hosts 
(for
> >>> example, to define where to add/remove the RPL IPv6 Hop-by-Hop 
Option if
> >>> used).
> >> 
> >>> To me, this interpretation of RPL Domain is thus only useful in a 
local
> >>> context and not to meant to designate one or more bounded set of 
nodes.
> >>> That's the role of DODAGs and Instances.
> >> 
> >>> Best Regards,
> >> 
> >>> Sébastien Dawans
> >> 
> >> On 11/16/2011 02:20 PM, Mukul Goyal wrote:
> >>> So, the revised doubts are as follows:
> >>> 
> >>> 1. It is clear that RPL routers are within an RPL domain but what 
about
> >>> the RPL-unaware IPv6 hosts attached to an RPL router? I would 
imagine
> >>> that such hosts are also within an RPL domain.
> >>> 
> >>> 2. Is an RPL domain same as an RPL instance? Or is an RPL domain a 
set
> >>> of RPL instances in the LLN? Can multiple RPL domains exist within 
an
> >>> LLN? Or is it that an RPL domain is same as an LLN using RPL as a
> >>> routing protocol?
> >>> 
> >>> THanks
> >>> Mukul
> >>> 
> >>> ----- Original Message -----
> >>> From: "Mukul Goyal"<[e-mail]>
> >>> To: "Thomas Heide Clausen"<[e-mail]>
> >>> Cc: "roll"<[e-mail]>
> >>> Sent: Wednesday, November 16, 2011 7:15:59 AM
> >>> Subject: Re: [Roll] definition of "RPL Domain"
> >>> 
> >>> 
> >>>> Now that we are at it: what is an RPL host? Or rather, why is this
> >>>> even a conceivable thing to define? Afaik, RPL is a routing 
protocol,
> >>>> and as such should concern only routers - not hosts?
> >>>> 
> >>> My bad. By RPL host, I actually meant an RPL leaf node. I think this
> >>> term "RPL host" was in use at one point in time but I cant find a
> >>> reference to it in current spec.
> >>> 
> >>> THanks
> >>> Mukul
> >>> 
> >>> 
> >>> 
> >>> ----- Original Message -----
> >>> From: "Thomas Heide Clausen"<[e-mail]>
> >>> To: "Mukul Goyal"<[e-mail]>
> >>> Cc: "JP Vasseur"<[e-mail]>, "roll"<[e-mail]>
> >>> Sent: Wednesday, November 16, 2011 6:25:31 AM
> >>> Subject: Re: [Roll] definition of "RPL Domain"
> >>> 
> >>> Now that we are at it: what is an RPL host? Or rather, why is this 
even
> >>> a conceivable thing to define? Afaik, RPL is a routing protocol, and 
as
> >>> such should concern only routers - not hosts?
> >>> 
> >>> I worry if this is inventing an entire ecosystem which "pretends to 
be
> >>> just like the Internet, except it is not", as it needs entirely new
> >>> stacks, protocols, translators/gateways everywhere, and with no real
> >>> traces of IP as we know it remaining?
> >>> 
> >>> 
> >> 
> >> _______________________________________________
> >> Roll mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/roll
> >> _______________________________________________
> >> Roll mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/roll
> > 
> > 
> > _______________________________________________
> > Roll mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/roll
> > _______________________________________________
> > Roll mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/roll
> _______________________________________________
> Roll mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/roll