Hi Anthony:

Thanks a whole bunch for this review. Let me please respond inline


[Anthony]
I went through your draft about the concept of a Backbone Router to help
large-scale networks 
deployments. I have a few comments about the draft, I put them below:

In Figure 1 you drew one single gateway that connects the backbone
router to the plant network. You do not refer to it later in the
document. Is it something application specific, a requirement for the
backbone router approach or a consequence to it? For any case, I would
mention it or remove it from the figure.

[Pascal]
The figure depicts the current ISA100.11a reference model. I need to add
words that this model applies to many situations in industrial, home and
building automation, but certainly not all and that the gateway could
become a firewall or a simple router depending on the use case. In the
same fashion, the plant network can be the building control network, the
Service Provider network or even the Internet.

[Anthony]
In Section 6.1, you wrote that Bacbone Routers announce themselves using
RAs that are broadcasted. Maybe you could see whether the solution
proposed in draft-chakrabarti-6lowpan-ipv6-nd-04.txt section 6.2 can
help avoid sending many broadcast messages. As a resume, they suggest:
" .. the PAN co-ordinator aka IPv6 router sends periodic RA to the
co-ordinators in its PAN by sending unicast messages to each of them...
Each co-ordinator can act as proxy IPv6 router advertiser and they can
broadcast the RA on behalf of the IPv6-router in their own domain
periodically.."

[Pascal] 
Very right. 

The focus of this doc is to illustrate the white board approach and the
use of proxy ND over the backbone to scale up the LoWPAN. If the group
agrees that this is the direction then we can dig further and merging
with draft-chakrabarti would be an excellent idea. It is clear to me
that my Draft 00 falls short in a number of occasions which were not the
focus at the time.

- registration mechanism. DHCP formats make more sense than NS/NA that I
used. I asked Benoit Lourdelet to join as a coauthor to rework that. I
also got comments that a totally new format could be welcome, along the
lines of RFC 3775 Binding Update (though the BU as it stands is not fit
for the job) but for the time being I'd rather stick to DHCPv6 and see
the feedback from the group.

- NS lookup. There can be 2 ways, proactive and reactive. My draft
reflects proactive using unicast NS/NA with the white board (and it will
be based on DHCP RFC 5007 in the next release). It's good because it
woorks for any type of address, link local, unique local and global.
draft-chakrabarti proposes a reactive version where the backbone router
does the lookup for you with an implicit request that comes with the
packet to be forwarded. This saves the NS/NA flow but causes all packets
to flow via the BbR. The route optimization with the PAN can still
happen under the BbR control using redirects. I think we need both
approaches make sense and are needed; that's a good reason to merge the
drafts.

- RA (as you point out), and that includes for me advertising the DHCP
relay and white board capabilities so we skip the SOLLICIT flow that is
normally multicast but makes little sense here. I had nothing much to
add to the excellent work that is already there in draft-chakrabarti;
that's another good reason to merge the drafts.
 

[Anthony]
In Section 6.4, you mention that the target link-layer address is that
of the destination if a shorcut is possible over the Lowpan. Do you plan
to specify the mechanism that will tell the nodes if they have to go
through the Backbone Router or not for a communication? Shortly, how do
the nodes know if a shorcut is possible and better than the backbone
route? Is it out of scope of this document?

[Pascal]

My draft does not say haw the routing is done (mesh under or route over)
and I do not think there is an assumption about it. What it does is a
proactive unicast NS/NA that results in the 16 or 64bit address of the
next hop. If the next hop is outside of the PAN then you get something
that leads to the BbR. If the nect hop is inside the PAN (the
destination) then you should get the address of that destination and you
packet will get there directly. 

There is a complexity involved in the case where multiple BbRs serve the
same PAN. They have to figure that the route inside the PAN exists and
is shorter than the route outside via the BbRs and the backbone. I think
that there's a point where we'll be needing ROLL to sort out these
issues. My draft does not cover that and expects a single BbR, hot
standby BbRs, or a default policy for inside routes vs. outside routes.

[Anthony]

There are various benefits of the Backbone router approach that may be
mentionned in the document. If you let backbone routers take care of the
routing for the nodes for instance, it may result in only unicast
packets within the different lowpans (towards and from the backbone
router), which will have as consequences 
1) less interferences (less packets in the air due to routes that are
known from the backbone router to the destination node and due to the
shorcut that the backbone may offer compared to long multihop
communications) 
2) less broadcast packets 
3) reduced memory requisite for routing tables on the nodes 
4) longer battery life for the sensor nodes. 
This could be made more visible in the document.


[Pascal]
This should be made more visible in the document, and though the
document does not specify how offline routing can be done there is a
common spirit between the two and the benefits are similar. 

Note that both ISA100.11a and WiHART assume offline routing and that it
will certainly be an option to consider for ROLL. I trust that most
readers figured that those benefits apply to this draft already but it
certainly does not hurt to add a small section about benefits that we
expect from avoiding multicast and using the backbone as we do.

Thanks a bunch Anthony,

Pascal
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan

Reply via email to