Re: [Roll] Comments on draft-dt-roll-p2p-rpl-01

Hi JP

Good starting point.

I would like to add that several application spaces have a need for
almost immediate delivery at any time.
For these applications, the optimal route is convenience while swift
delivery is a must. Reactive discovery
is the way to guarantee swift delivery at any time.

Thanks,
  Anders

________________________________

	From: roll-bounces@ietf.org [mailto:roll-bounces@ietf.org] On
Behalf Of JP Vasseur
	Sent: Wednesday, June 16, 2010 12:11
	To: Mukul Goyal; Emmanuel Baccelli
	Cc: ROLL WG
	Subject: Re: [Roll] Comments on draft-dt-roll-p2p-rpl-01

	Hi, 

	As promised here is some text wrt the trade-off on on-demand
routing.

	"As pointed out in this document, RPL does provide support of
P2P routes via a common
	ancestor. Using the mechanism described in this document, more
optimal routes with 
	regards to some metrics may be discovered but it is worth
pointing out that this comes
	at the price of additional control messages in the network since
the discovery process
	requires sending additional routing control plane messages. The
gain in terms of path 
	optimality may vary with the network topology and networking
conditions. Thus network 
	designer may want to take into account the trade-off between the
potential discovery of a 
	more optimal route and the associated cost in terms of network
and node resources.
	Note that on-demand routing also offers the benefit of not
having to maintains states 
	as with proactive routing."

	Thanks.

	JP.

	On Jun 16, 2010, at 12:00 PM, JP Vasseur wrote:

		Dear authors,

		Since I had a number of comments I inserted them in the
ID itself. As you will notice:
		1) Some of them are purely editorial
		2) Others are more fundamental and may require just
clarifications, more discussion on the mailing list, ... In this case,
feel free to start a separate thread (note that ticket cannot be
assigned to non WG document).
		3) I also suggested new sections

		Internet Engineering Task Force
M. Goyal, Ed.
		Internet-Draft                         University of
Wisconsin Milwaukee
		Intended status: Informational
E. Baccelli, Ed.

		JP> Are you sure that you want to make it informational
?

		Expires: December 16, 2010
INRIA

P2P. Team

June 14, 2010

		   Reactive Discovery of Point-to-Point Routes in Low
Power and Lossy
		                                Networks
		                        draft-dt-roll-p2p-rpl-01

		Abstract

		   This document describes a mechanism to discover and
establish "on
		   demand" one or more routes between two routers in a
low power and
		   lossy network.

		JP> Needs a bit more details + explicitly refer to IPv6
in the abstract 
		and introduction.
		Status of this Memo This Internet-Draft is submitted to
IETF in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task
Force (IETF). Note that other groups may also distribute working
documents as Internet-Drafts. The list of current Internet- Drafts is at
http://datatracker.ietf.org/drafts/current/. Internet-Drafts are draft
documents valid for a maximum of six months and may be updated,
replaced, or obsoleted by other documents at any time. It is
inappropriate to use Internet-Drafts as reference material or to cite
them other than as "work in progress." This Internet-Draft will expire
on December 16, 2010. Copyright Notice Copyright (c) 2010 IETF Trust and
the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions
Relating to IETF Documents (http://trustee.ietf.org/license-info) in
effect on the date of publication of this document. Please review these
documents carefully, as they describe your rights and restrictions with
respect to this document. Code Components extracted from this document
must include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as Goyal,
et al. Expires December 16, 2010 [Page 1]  Internet-Draft
draft-dt-roll-p2p-rpl-01 June 2010 described in the Simplified BSD
License. Table of Contents 1. Introduction . . . . . . . . . . . . . . .
. . . . . . . . . . 3 2. Targeted Use Cases . . . . . . . . . . . . . .
. . . . . . . . 4 3. Terminology . . . . . . . . . . . . . . . . . . . .
. . . . . 4 4. Functional Overview . . . . . . . . . . . . . . . . . . .
. . 5 5. Propagation of Discovery Messages . . . . . . . . . . . . . . 7
5.1. The Route Discovery Option . . . . . . . . . . . . . . . . 7 5.2.
Setting a DIO Carrying a Route Discovery Option . . . . . 9 5.3.
Processing of a DIO Carrying a Route Discovery Option At An Intermediate
Node . . . . . . . . . . . . . . . . . 10 5.4. Processing of a DIO
Carrying a Route Discovery Option At The Target Router . . . . . . . . .
. . . . . . . . . . 12 6. Propagation of Discovery Reply Messages . . .
. . . . . . . . 13 6.1. The Discovery Reply Object (DRO) . . . . . . . .
. . . . . 13 6.1.1. The Source Route Option . . . . . . . . . . . . . .
. 15 6.2. DRO as Acknowledgement for Backward Source Routes . . . . 16
6.3. DRO as Carrier of Forward/Bidirectional Source Routes . . 17 6.4.
Establishing Hop-by-hop Routes Via DRO . . . . . . . . . . 17 7.
Security Considerations . . . . . . . . . . . . . . . . . . . 18 8. IANA
Considerations . . . . . . . . . . . . . . . . . . . . . 18 9. Authors
and Contributors . . . . . . . . . . . . . . . . . . . 18 10. References
. . . . . . . . . . . . . . . . . . . . . . . . . . 18 10.1. Normative
References . . . . . . . . . . . . . . . . . . . 18 10.2. Informative
References . . . . . . . . . . . . . . . . . . 18 Authors' Addresses . .
. . . . . . . . . . . . . . . . . . . . . . 19 Goyal, et al. Expires
December 16, 2010 [Page 2]  Internet-Draft draft-dt-roll-p2p-rpl-01
June 2010 1. Introduction RPL [I-D.ietf-roll-rpl] provides
multipoint-to-point (MP2P) routes from routers in a low power and lossy
network (LLN) to a sink router by organizing the routers along a
Directed Acyclic Graph (DAG) rooted at the sink.  

		JP> low power and lossy .... Low Power and Lossy

		JP> RPL provides P2MP, MP2P and P2P routes too.

		The DAG root initiates the DAG formation by originating
		   a DODAG Information Object (DIO) message.  The DIO
message is sent
		   via link-local multicast and carries information
about the
		   originating router's position (the "rank") in the
DAG.  On receiving
		   DIO messages sent by its neighbors, a router
determines its own
		   "rank" in the DAG and originates its own DIO message.

		JP> Without going too much in the details, you may just
want to add
		that a node decides to join a DODAG according to a
number of parameters
		(identity of the DODAG, OF, routing/constraints a
specified in the metric-ID,...).
		   RPL enables point-to-multipoint (P2MP) routing from a
router to its
		   descendants in the DAG by allowing a router to send a
Destination
		   Advertisement Object (DAO) upwards along the DAG.
The DAO carries
		   the aggregated 

		JP> Potentially "aggregated"
		information regarding the descendants (and other local
		   prefixes) reachable through the originating router.

		   RPL also provides mechanisms for point-to-point (P2P)
routing between
		   any two routers in the DAG.  

		JP> Yes.
		If the destination is within the
		   source's "range", the source may directly send
packets to the
		   destination.  Otherwise, if the destination is a DAG
descendant and
		   the source maintains "downwards" routing state about
this descendant,
		   it can forward the packet along this route.
Otherwise, the source
		   sends the packet to a DAG parent, which then applies
the same set of
		   rules to forward the packet further.  Thus, a packet
travels up the
		   DAG until it reaches a router that knows of the
downwards route to
		   the destination and then it travels down the DAG
towards its
		   destination.  A router may or may not maintain
routing state about a
		   descendant depending on whether its immediate
children send it such
		   information in their DAOs and whether the router can
store this
		   information.  Thus, in the best case scenario, the
"upwards" segment
		   of the P2P route between a source and a destination
ends at the first
		   common ancestor of the source and the destination.
In the worst
		   case, the "upwards" segment would extend all the way
to the DAG's
		   root.  If the destination did not originate a DAO,
the packet will
		   travel all the way to the DAG's root, where it will
be dropped.

		JP> This may require to be a bit more accurate here. If
there is no DAO,
		there RPL is used for MP2P routing only (that is not a
protocol design 
		characteristic though). 
		   The P2P routing functionality available in RPL
suffers from several
		   shortcomings
[I-D.brandt-roll-rpl-applicability-home-building]:

		JP> "suffering from several shortcomings" is a bit
strong though. You may
		want to say that additional mechanisms may be required
to address specific
		situation and list them. 
		   o  The need to maintain routes "proactively", i.e.
every possible
		      destination in the DAG must originate a DAO.

		   o  The constraint to route only along a DAG
potentialy leading to
		      significantly suboptimal P2P routes and traffic
congestion near
		      the DAG root.

		JP> 
		First bullet: you can mention that this requires to
pro-actively advertise the
		route and to keep refreshing the state accordingly
		Second bullet: just to be very accurate, mention that
the route may be sub-optimal.
		This is highly network topology / OF dependent.
		Goyal, et al. Expires December 16, 2010 [Page 3] 
Internet-Draft draft-dt-roll-p2p-rpl-01 June 2010 These shortcomings are
not compatible with the home and commercial building domain application
requirements described in [RFC5826] and
[I-D.ietf-roll-building-routing-reqs].  

		JP> Again, to be accurate, there are no MUST
requirements listed in
		these ID that are not satisfied by RPL (with regards to
the aforementioned
		issues). I am NOT saying that the proposed mechanism is
not useful 
		(by all means), just that the statement above is not
correct.

		Such applications require a
		   mechanism providing source-initiated discovery of P2P
routes that are
		   not along a DAG.  

		JP> In some cases, it may be desirable to make use of
source-initiated ... 

		This document thus describes such a mechanism,
		   complementary to the basic RPL specification.

		   The specified scheme is based on a reactive on-demand
approach, which
		   enables a router to discover one or more
"good-enough" routes in
		   either direction between itself and another router in
the LLN without
		   any constraints regarding the existing DAG-membership
of the links
		   that such routes may use.  

		JP> This will be an aspect triggering some discussion.
How "good" is
		"good enough" ? You will need some text to elaborate
here or point 
		to another section explaining what is meant by "good
enough"
		Such routes may be source-routes or hop-
		   by-hop ones.  

		JP> Just terminology: instead of referring to hop-by-hop
routes, you may
		want to say that once a route has been computed between
two nodes in 
		the network, IPv6 packet may be forward using a source
routing mechanism
		as specified in [XX] or in a hop by hop fashion.
		A complementary functionality, necessary to help decide
		   whether to initiate a route discovery, is a mechanism
to measure the
		   end-to-end cost of an existing route.  

		JP> s/cost/path cost
		Such functionality will be
		   described in a separate document.

		JP> But its use is described in this document ?
		2. Targeted Use Cases The mechanisms described in this
document are intended to be employed as complementary to RPL in specific
scenarios that need point-to- point (P2P) routes between arbitrary
routers. One target use case, common in a home environment, involves a
remote control (or a motion sensor) that suddenly needs to communicate
with a lamp module, whose network address it knows apriori. In this
case, the source of data (the remote control or the motion sensor) must
be able to discover a route to the destination (the lamp module) "on
demand". Another target use case, common in a large commercial building
environment, involves a large LLN deployment where P2P communication
along a particular DAG among hundreds (or thousands) of routers creates
severe traffic congestion near that DAG's root, and thus routes across
this DAG are desirable. Targeted use cases also include scenarios where
energy or delay constraints are not satisfied by the P2P routes along a
DAG because they involve traversing many more intermediate routers than
necessary to reach the destination. 

		JP> What I would suggest is to summarize it, put in the
introduction,
		and detail the use case in the applicability statement.
		3.  Terminology

		   The key words "MUST", "MUST NOT", "REQUIRED",
"SHALL", "SHALL NOT",
		   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT
RECOMMENDED", "MAY", and
		   "OPTIONAL" in this document are to be interpreted as
described in

		Goyal, et al.           Expires December 16, 2010
[Page 4]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   [RFC2119].

		   Additionally, this document uses terminology from
		   [I-D.ietf-roll-terminology] and [I-D.ietf-roll-rpl],
and introduces
		   the following terminology:

		   Origin Router: The router initiating the route
discovery.  The origin
		   router acts as one end point of the routes to be
discovered.

		   Target Router: The other end point of the routes to
be discovered.

		   Intermediate Router: A router that is neither the
origin nor the
		   target.

		   Forward Route: A route from the origin router to the
target router.

		   Backward Route: A route from the target router to the
origin router.

		   Bidirectional Route: A route that can be used in both
directions:
		   from the origin router to the target router and vice
versa.

		   Source Route: A complete and ordered list of routers
that can be used
		   by a packet to travel from a source to a destination.
Such source
		   routes can be carried by a packet in a proposed Type
4 Routing Header
		   [I-D.hui-6man-rpl-routing-header].

		   Hop-by-hop Route: A route from a source to a
destination where each
		   router in the route knows only the address of the
next hop on the
		   route.

		JP> I would suggest to change the definition for
"routing paradigm
		whereby packets are routed from their source to
destination by 
		on a hop by hop basis where the packets is routed by a
set of routers
		along the path according to the routing table stored by
each router
		along the path".
		   In this document, the term 'router' refers to any LLN
device that can
		   forward packets generated at other devices.

		4.  Functional Overview

		   Router A originates a "Discovery" message listing
router B as the
		   target.  Node A also indicates in the message:

		   o  The nature of the routing cost, the method for
accumulating the
		      end-to-end cost 

		JP> This may de detailed hereafter but I guess that you
will point to 
		the metric-ID for the metric used?

		and the criteria used to determine if a route is
		      "good enough";

		   o  The direction (forward: router A to router B;
backward: router B
		      to router A; or both) of the route being
discovered;

		   o  The desired number of routes;

		Goyal, et al.           Expires December 16, 2010
[Page 5]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   o  Whether the route is a source-route or a
hop-by-hop one; and

		   o  A limit on the "distance" the Discovery message
may travel.

		JP> Need to define the term "distance": is it a max path
cost, number of
		hops, ... ?
		The Discovery message propagates via link-local
multicast 

		JP> Indicate "IPv6"
		with each
		   receiving router making the decision regarding
whether to forward the
		   message further based on the "distance" (from router
A) that this
		   particular copy of the message has already traveled.
Note that this
		   document does not require a receiving router to use
the "good enough"
		   criteria to make the forwarding decision.  This is
because the
		   evaluation of such criteria may be too complex for a
constrained
		   intermediate router to perform for each received copy
of the
		   Discovery message.  

		JP> So I guess that you define later what to do in both
cases (the node
		cannot process the "good enough" criteria or it can
process it) ?
		The calculation of the "distance" that a copy of
		   the Discovery message has already travelled should be
simple enough
		   for a constrained router to perform.  

		JP> Again it is hard to tell from what you wrote if this
can reasonably be
		achieved; it depends on the definition of the term
"distance".
		A router may optionally decide
		   not to forward a copy of the Discovery message
further because the
		   aggregated cost of the route so far fails the "good
enough" criteria.

		JP> You will nee to be more specific here. Is it a "may"
or a "MAY". 
		Define what happens in this case. You may provide more
details later
		in the document.
		   The underlying mechanism being used to propagate the
Discovery
		   message may put further restrictions on its
propagation.  A router
		   should not propagate the Discovery message further if
hop-by-hop
		   routes are desired and the router can not store state
for this route.

		   As a copy of the Discovery message travels towards
router B, it
		   accumulates the relevant forward/backward costs as
well as the route
		   it takes.  When router B receives a copy of the
Discovery message, it
		   determines whether the route traveled by the message
meets the "good
		   enough" criteria.

		JP> I think that you already mentioned this.
		   If router A had requested the discovery of backward
source-routes,

		JP> Didn't you mean "forward" ?
		   router B caches one or more good enough source-routes
it identifies.
		   Additionally, router B sends one or more "Discovery
Reply" message to
		   router A to acknowledge the discovery of these
routes.  These
		   acknowledgements allow router A to judge the success
of the route
		   discovery.  The Discovery Reply messages can travel
towards router A
		   in any manner chosen by router B. For example, router
B may source-
		   route the messages or send them towards router A
along a DAG.

		   If router A had requested the discovery of "n"
forward source-routes,
		   router B sends the "n" good enough source-routes it
identifies to
		   router A in one or more Discovery Reply messages.
Again, these
		   Discovery Reply messages can travel towards router A
in any manner
		   chosen by router B.

		JP> What if the DAG is "broken": retry from A after
expiration of a local
		timer ?
		   If router A had requested the discovery of "n"
bidirectional source-
		   routes, router B caches the "n" good enough
source-routes it
		   identifies and also sends these routes to router A in
one or more
		   Discovery Reply messages.  These Discovery Reply
messages can travel
		   towards router A in any manner chosen by router B.

		Goyal, et al.           Expires December 16, 2010
[Page 6]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   If router A had requested the discovery of "n"
forward/backward/
		   bidirectional hop-by-hop routes, router B sends out a
Discovery Reply
		   message to router A for each one of the "n" good
enough routes it
		   identifies.  

		JP> Do you need one "Discovery Reply message" per route
or next-hop
		in this case?
		The Discovery Reply message travels towards router A
		   using the source-route accumulated by the Discovery
message.  

		JP> In the case of hop-by-hop routes, source routes,
both ?
		As this
		   message travels towards router A, it establishes
appropriate forward/
		   backward routing state in the en-route routers.  This
"non DAG"
		   routing state should be specially marked to associate
it with the
		   routing metrics 

		JP> Refer to metric ID ?
		used to discover the route and to distinguish such
		   state from other DAG-specific routing state the
router may have.

		JP> Not sure that you need to distinguish the state.
Suppose that you
		discover a hop-by-hop route (not a big fan of that terms
as pointed out
		above but let's continue to use it for the time being),
a storing node
		may indeed have two routes for the same destination: do
you want to
		state that it should always use the P2P routes
discovered by this
		mechanism or let it choose the best one according to the
metric in
		use?

		JP> What if there are two routes using different
metrics?
		5.  Propagation of Discovery Messages

		   RPL uses DIO message propagation to build a DAG.  The
DIO message
		   travels via link-local multicast.  

		JP> Add everywhere "IPv6 link-local multicast addresses"

		Each router joining the DAG
		   determines a rank for itself and ignores the
subsequent DIO messages
		   received from lower (higher in numerical value)
ranked neighbors.
		   Thus, the DIO messages propagate outwards rather than
return inwards
		   towards the DAG root.  

		JP> I would suggest to align to the RPL terminology.
Inward/outward
		are not used anymore.

		The DIO message generation at a router is
		   further controlled by a trickle timer that allows a
router to avoid
		   generating unnecessary messages.  

		JP> Add a reference to the Trickle ID

		The link-local multicast based
		   propagation, trickle-controlled generation and the
rank-based
		   poisoning of messages traveling in the wrong
direction (towards the
		   DAG root) provide powerful incentives to use the DIO
message as the
		   Discovery message and propagate the DIO/Discovery
message by creating
		   a "temporary" DAG.

		5.1.  The Route Discovery Option

		       0                   1                   2
3
		        0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3
4 5 6 7 8 9 0 1

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       |   Type = 9    | Option Length | D |S|  N  | L
|O|  Max Rank   |

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       |
|
		       |                       Target Address
|
		       |
|
		       |
|

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       |              OCP              |
		       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

		              Figure 1: Format of the Route Discovery
Option

		   In order to be used as a Discovery message, a DIO
MUST carry a "Route
		   Discovery" option illustrated in Figure 1.  A DIO
MUST NOT carry more

		Goyal, et al.           Expires December 16, 2010
[Page 7]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   than one Route Discovery options.  

		JP> Explain what you do if more than one is present:
			* Ignore the second one?
			* Ignore the entire message?

		A Route Discovery option consists
		   of the following fields:

		   o  Option Type = 0x09 (to be confirmed by IANA).

		   o  Option Length = 20 or 22 octets depending on
whether the OCP field
		      is included or not.

		   o  D: A 2-bit field that indicates the direction of
the desired
		      routes:

		      *  D = 0x00: Forward;

		      *  D = 0x01: Backward;

		      *  D = 0x02: Bidirectional.

		   o  S: A flag that indicates whether source routes are
desired.  The
		      flag is set if the source routes are desired.  The
flag is clear
		      if hop-by-hop routes are desired.

		   o  N: A 3-bit unsigned integer indicating the number
of routes
		      desired.

		   o  L: A 2-bit field indicating the minimum "Life
Time" of the
		      temporary DAG, i.e., the minimum duration a router
joining the
		      temporary DAG must maintain its membership in the
DAG:

		      *  L = 0x00: Minimum life time is 5 seconds;

		      *  L = 0x01: Minimum life time is 10 seconds;

		      *  L = 0x02: Minimum life time is 1 minute;

		      *  L = 0x03: Minimum life time is 10 minutes.

		JP> Placeholder ... indicate what happens if a node
cannot store that
		additional state (add a manageability section).
		   o  O: A flag that indicates whether the same OCP is
used for route
		      discovery as well as temporary DAG formation.  If
this flag is
		      clear, the OCP contained in the DODAG
Configuration option in the
		      DIO is used for route discovery as well.
Otherwise, the OCP
		      specified in the Route Discovery option is used
for route
		      selection 

		JP> s/route selection/route discovery

		and the metrics/constraints used for route selection are
		      carried in a Metrics Container option immediately
following the
		      Route Discovery option.

		JP> Since OF is decoupled from metrics, you always need
to indicate
		the metric in use. That said, I do not see your point
here. Indeed, the 
		temporary DAG is built according to some OF and metrics
so the 
		resulting routes will inherit these metrics?
		   o  Max Rank: A 7-bit unsigned integer that indicates
the maximum rank
		      allowed in the temporary DAG advertised by the DIO
message.  This
		      upper limit on the DAG rank serves as the
"distance" referred to

		Goyal, et al.           Expires December 16, 2010
[Page 8]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		      in Section 4 and provides a control over how far
the Route
		      Discovery option contained in the DODAG
Configuration option in
		      the DIO message may travel.  

		JP> You may discuss it later in this document ... but
the distance is tied
		to how you compute the rank.
		A router MUST not 

		JP> s/MUST not/MUST NOT
		join the temporary
		      DAG if its rank in the DAG will exceed this limit.
Later versions
		      of this draft will map the 128 value space
available in this field
		      to 16-bit long limits on the DAG rank.

		   o  Target Address: The IPv6 address of the target
router.

		   o  OCP: 16 bit unsigned integer.  An optional field,
present only if
		      the O flag is set, that indicates the objective
code point (OCP)
		      to be used for route selection.

		5.2.  Setting a DIO Carrying a Route Discovery Option

		   A DIO message that carries a Route Discovery option
MUST set the Base
		   Object, described in [I-D.ietf-roll-rpl], in the
following manner:

		   o  RPLInstanceID: RPLInstanceID MUST be a local value
as described in
		      Section 4.1 of [I-D.ietf-roll-rpl].

		   o  Grounded (G) Flag: MUST be clear since the
objective of DAG
		      formation is the propagation of Route Discovery
option.  This DAG
		      is temporary in nature and is not used for routing
purpose.

		   o  Destination Advertisement Supported (A) Flag: MUST
be clear for
		      same reasons as described above.

		   o  Destination Advertisement Trigger (T) Flag: MUST
be clear.

		   o  Mode of Operation (MOP): Since the temporary DAG
is not to be used
		      for routing purposes, the value of this field is
immaterial.  To
		      allow constrained routers to join the DAG, the MOP
field SHOULD be
		      set to 00 (non-storing).

		JP> I'd rather suggest to specify a new MOP value (0x02)
		   o  DODAGPreference (Prf): TBD

		   o  Destination Advertisement Trigger Sequence Number
(DTSN): TBD

		JP> To be discussed on the mailing list ?
		   o  DODAGID: IPv6 address of the router initiating the
route
		      discovery.

		JP> Origin router.
		   The other fields in the Base Object are set as per
the rules
		   described in [I-D.ietf-roll-rpl].

		   The DODAG Configuration option, carried in the DIO
message, specifies
		   the parameters for the trickle timer operation that
governs the
		   generation of DIO messages by the routers joining the
temporary DAG.

		JP> Also according to RPL ID or would you recommend
different
		default values ? Default values for RPL are TBD but we
could easily
		think of different values in this case. To be revisited
in the future.

		Goyal, et al.           Expires December 16, 2010
[Page 9]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   The other fields defined in the DODAG Configuration
option are set as
		   follows:

		   o  The MaxRankIncrease field MUST be set to 0 to
disable local repair
		      of the temporary DAG.

		   o  This document RECOMMENDS a value 1 for the
MinHopRankInc field.

		   o  Objective Code Point (OCP): The OCP to be used for
temporary DAG
		      formation.  This document RECOMMENDS RPL Objective
Function 0, as
		      defined in [I-D.ietf-roll-of0], for use as the
objective function
		      for the formation of the temporary DAG.

		   A DIO, that contains a Route Discovery option with O
flag set, MUST
		   also contain a Metric Container option immediately
following the
		   Route Discovery option.  This Metric Container option
carries the
		   values for the routing metrics as well as the
constraints
		   (constituting the "good enough" criteria) used for
route selection.

		JP> This is confusing: the metric and constraint carried
within the Metric
		Container are used for the DODAG formation. Do you want
to use them
		for a different purpose ? At this point, you have not
yet defined the use of
		the good enough criteria but one can guess that you
would use the good
		enough metrics to control the flooding and the metric/OF
to effectively 
		build the temporary DAG and discover the routes. If so,
the good enough
		metrics are of different use.
		   Note that if O flag in the Route Discovery option is
clear, the OCP
		   and Metric Container used for temporary DAG formation
are used for
		   route selection as well.

		   A DIO, carrying a Route Discovery option, MUST not
carry any Route

		JP> s/MUST not/MUST NOT
		   Information or Prefix Information options described
in
		   [I-D.ietf-roll-rpl].

		5.3.  Processing of a DIO Carrying a Route Discovery
Option At An
		      Intermediate Node

		   The rules for DIO processing and transmission,
described in Section 7
		   of RPL [I-D.ietf-roll-rpl], apply to DIOs carrying a
Route Discovery
		   option as well except as modified in this document.

		   When an intermediate router joins a temporary DAG
advertized by a DIO
		   carrying the Route Discovery option, it MUST maintain
its membership
		   in the DAG for the Minimum Life Time duration listed
in the Route
		   Discovery option.  

		JP> What if it cannot ? Just not join the DAG and log an
error ?
		Maintaining membership in the DAG implies
		   remembering:

		   o  The RPLInstanceID, the DODAGID and the
DODAGVersionNumber for the
		      temporary DAG;

		   o  The router's rank in the temporary DAG;

		   o  The best Metric Container, 

		JP> you mean the best path cost?
		along with the associated source route
		      from the initiator of route discovery till this
router (carried in
		      a Record Route IPv6 Extension Header proposed in
		      [I-D.thubert-6man-reverse-routing-header]), for
the Route

		Goyal, et al.           Expires December 16, 2010
[Page 10]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		      Discovery option being propagated by the temporary
DAG.  If the
		      router can not compare Metric Containers, it MUST
remember the
		      latest Metric Container it has received, along
with the associated
		      source route.

		JP> Requires clarification on what you mean by "compare"
		   A router belonging to a temporary DAG need not
remember the identity
		   of its DAG parents since the temporary DAG is not
used for routing.

		JP> Are you sure ? What is a hop-by-hop route is needed
? Don't you
		need to remember your parent, or do you still record the
route ?
		   The main purpose of a temporary DAG's existence is to
facilitate the
		   propagation of the Route Discovery option.

		   The router does not process a Route Discovery option,
contained in
		   the received DIO, any further if any of the following
conditions are
		   true:

		   o  The router does not support the objective function

		JP> and/or metrics
		being used for
		      route discovery

		   o  The router does not have sufficient information to
calculate the
		      relevant routing metrics/constraints in the right
direction (as
		      specified by the D field).

		   o  The S field is clear, i.e. hop-by-hop routes are
desired, but the
		      router can not participate in a hop-by-hop route.

		   o  The router is an intermediate router and the
router's rank in the
		      temporary DAG, calculated on the basis of the rank
and objective
		      function carried in the Base Object in the DIO,
exceeds the Max
		      Rank value specified in the Route Discovery
option.

		   A router MUST discard the DIO if the Route Discovery
option contained
		   in the message does not need further processing.  

		JP> Log an error ?
		Otherwise, the
		   Route Discovery option is processed as follows.

		   The router updates the Metrics Container associated
with the received
		   Route Discovery option as per the specified objective
function.  

		JP> And routing metrics

		The
		   router may optionally check the Metric Container to
determine if the
		   aggregated values for the routing metrics meet all
the constraints
		   listed in the Metric Container.  

		JP> The aggregated values are for the metrics.
Constraints are used to
		prune potential path.
		If not, the Route Discovery option
		   is discarded 

		JP> You mean the DIO message.
		without further processing.  Otherwise, the router
		   updates its in-memory copy of the latest/best Metric
Container for
		   this Route Discovery option along with the associated
source route
		   updated in the correct direction (based on the D
field of the Route
		   Discovery option).  Any change in the in-memory copy
also requires
		   resetting the trickle timer and generating a new DIO
carrying the
		   Route Discovery option, the updated latest/best
Metric Container and
		   the associated source route (in a Record Route IPv6
extension header
		   proposed in
[I-D.thubert-6man-reverse-routing-header]) when the timer
		   fires.  

		JP> If the changes occur during the lifetime window.
		Note that the Metric Container MUST immediately follow
the

		Goyal, et al.           Expires December 16, 2010
[Page 11]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   Route Discovery option if the O flag in the Route
Discovery option is
		   set.

		5.4.  Processing of a DIO Carrying a Route Discovery
Option At The
		      Target Router

		   When the target router joins the temporary DAG
advertized by a DIO
		   carrying the Route Discovery option, it MUST maintain
its membership
		   in the DAG for the Minimum Life Time duration listed
in the Route
		   Discovery option.  Maintaining membership in the DAG
implies
		   remembering:

		   o  The RPLInstanceID, the DODAGID and the
DODAGVersionNumber for the
		      temporary DAG;

		   o  The router's rank in the temporary DAG;

		   The target router does not process a Route Discovery
option,
		   contained in the received DIO, any further if any of
the following
		   conditions are true:

		   o  The router does not support the objective function
used for route
		      discovery

		   o  The router does not have sufficient information to
calculate the
		      relevant routing metrics/constraints in the right
direction (as
		      specified by the D field).

		JP> or does not support the metric.
		The target router MUST discard the DIO if the Route
Discovery option contained in the message does not need further
processing.

		JP> ? How does it knows (simply because it reaches the
destination) ?
		   Otherwise, the Route Discovery option is processed as
follows.

		   The target router updates the Metrics Container
associated with the
		   received Route Discovery option as per the specified
objective
		   function.  

		JP> and routing metrics (remember they are decoupled).

		The target router then checks the Metric Container to
		   determine if the aggregated values for the routing
metrics meet all
		   the constraints listed in the Metric Container.  

		JP> See previous comments.

		If not, the Route
		   Discovery option is discarded without further
processing.  Otherwise,
		   the router MAY select the source route accumulated by
the received
		   DIO as one of the discovered routes.  

		JP> "MAY" ? When ?
		The target router MUST send one
		   or more RPL Control Messages carrying a Discovery
Reply Object
		   (defined in the next section) back to the origin
router (identified
		   by the DODAGID field in the DIO Base Object) as
discussed in the
		   following sections.

		   The target router MUST NOT forward a DIO carrying a
Route Discovery
		   option that lists one of its own addresses as the
Target Address.

		Goyal, et al.           Expires December 16, 2010
[Page 12]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		6.  Propagation of Discovery Reply Messages

		6.1.  The Discovery Reply Object (DRO)

		       0                   1                   2
3
		        0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3
4 5 6 7 8 9 0 1

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       | RPLInstanceID |    Version    | D |S|  N  |
Reserved      |

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       |
|
		       |                         DODAGID
|
		       |
|
		       |
|

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       |
|
		       |                       Target Address
|
		       |
|
		       |
|

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
		       | Option(s)...
		       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+...

		           Figure 2: Format of the Discovery Reply
Object (DRO)

		   This document defines a new RPL Control Message type,
the Discovery
		   Reply Object (DRO) with code 0x04 (to be confirmed by
IANA), that
		   serves the following functions:

		   o  An acknowledgement from the target router to the
origin router
		      regarding the successful discovery of backward
source routes;

		JP> ACK of routes from the origin to the target: you may
not have routes
		in the opposite direction.
		   o  Carries one or more forward/bidirectional source
routes from the
		      target to the origin router;

		   o  Establishes a hop-by-hop
forward/backward/bidirectional route as
		      it travels from the target to the origin router.

		   The format for a Discovery Reply Object (DRO) is
shown in Figure 2.
		   A DRO consists of the following fields:

		   o  RPLInstanceID: The RPLInstanceID of the temporary
DAG used for
		      route discovery.

		   o  Version: The Version of the temporary DAG used for
route
		      discovery.

		Goyal, et al.           Expires December 16, 2010
[Page 13]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		   o  DODAGID: The DODAGID of the temporary DAG used for
route
		      discovery.  The DODAGID also identifies the origin
router.  The
		      RPLInstanceID, the Version and the DODAGID
together uniquely
		      identify the temporary DAG used for route
discovery and can be
		      copied from the Base Object of the DIO advertizing
the temporary
		      DAG.

		   o  D: A 2-bit field that indicates the direction of
the discovered
		      routes:

		      *  D = 0x00: Forward;

		      *  D = 0x01: Backward;

		      *  D = 0x02: Bidirectional.

		      This field has the same value as the corresponding
field in the
		      Route Discovery option.

		   o  S: A flag that is clear if the Discovery Reply
Object is
		      establishing an hop-by-hop route.  The flag is set
if the
		      Discovery Reply Object carries (or is an
acknowledgement for the
		      discovery of) one or more source routes.

		   o  N: A 3-bit field that indicates the number of
source routes
		      carried or acknowledged in the Discovery Reply
Object.

		   o  Target Address: The IPv6 address of the target
router originating
		      the Discovery Reply Object.

		   o  Reserved: These bits are reserved for future use.
These bits MUST
		      be set to zero on transmission and MUST be ignored
on reception.

		   o  Options: The Discovery Reply Object MAY carry up
to 7 Source Route
		      options (defined in the next section) with each
such option
		      carrying a complete forward/bidirectional source
route and
		      optionally followed by a Metric Container option
that lists the
		      aggregated values for the routing metrics for the
source route.

		JP> This may not be an aggregated value but recorded
metric too.
		Goyal, et al. Expires December 16, 2010 [Page 14] 
Internet-Draft draft-dt-roll-p2p-rpl-01 June 2010 6.1.1. The Source
Route Option 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type
= 10 | Option Length | Compr | Pad | Resvd |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | .
Addresses[1..n] . . . | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure
3: Format of the Source Route Option The Source Route option,
illustrated in Figure 3, carries a complete forward/bidirectional source
route from the target router to the origin router. A Source Route option
can only be a part of the Discovery Reply Object and MAY be immediately
followed by a Metric Container option that contains the aggregated
values of the routing metrics for this source route. A Route Discovery
option consists of the following fields: o Option Type = 0x0A (to be
confirmed by IANA). o Option Length = Variable, depending on the size of
the Addresses vector. o Compr: 4-bit unsigned interger indicating the
number of prefix octets that are elided from each address. For example,
Compr value will be 0 if full IPv6 addresses are carried in the
Addresses vector. o Pad: 4-bit unsigned integer. Number of octets that
are used for padding between Address[n] and the end of the Source Route
option. o Address[1..n]: Vector of addresses, numbered 1 to n. Each
vector element has size (16 - Compr) octets. A common network
configuration for an RPL domain is that all routers within an LLN share
a common prefix. The Source Route option uses the Compr field to allow
compaction of the Addresses[1..n] vector when all entries share the same
prefix as the DODAGID or the Target Address of the encapsulating
Discovery Reply Object. The shared prefix octets are not carried within
the Source Route option and each entry in Address[1..n] has size (16 -
Compr) octets. When Compr is Goyal, et al. Expires December 16, 2010
[Page 15]  Internet-Draft draft-dt-roll-p2p-rpl-01 June 2010 non-zero,
there may exist unused octets between the last entry, Address[n], and
the end of the Source Route option. The Pad field indicates the number
of unused octets that are used for padding. Note that when Compr is 0,
Pad MUST be null and carry a value 0. The Source Route option MUST NOT
specify a path that visits a router more than once. When generating a
Source Route option, the target router may not know the mapping between
IPv6 addresses and routers. Minimally, the target router MUST ensure
that: o The IPv6 Addresses do not appear more than once; o The IPv6
addresses of the origin and the target routers do not appear in the
Address vector; o The Address vector represents a source route in
forward direction with Address[1] being the next hop for the origin
router. Multicast addresses MUST NOT appear in a Source Route option. 

		JP> you may want to add a ref to RH4 since the format is
very similar.
		6.2.  DRO as Acknowledgement for Backward Source Routes

		   When a target router discovers a backward source
route, 

		JP> I have an issue with your terminology. When the
target receives the
		DIO with route discovery option, what is discovered is a
forward route ?

		it sends a
		   DRO message to the origin router as an acknowlegement
for the
		   discovered route.  Optionally, a target router MAY
send a DRO
		   acknowledgement to the origin router after
discovering multiple
		   backward source routes.  A DRO, serving as an
acknowledgement for
		   backward source route discovery, has its D field set
to 0x01
		   (indicating backward) while the S flag is set to 1
(indicating source
		   route).  The N field is set to indicate the number of
discovered
		   backward source routes being acknowledged.  Such a
DRO message MUST
		   NOT contain any option.

		   This document does not require a particular method
for sending such a
		   DRO message to the origin router.  The target router
MAY send the DRO
		   message to the origin router in any fashion,
including:

		   o  Using a source route carried in a Type 4 Routing
header
		      [I-D.hui-6man-rpl-routing-header];

		   o  Along any DAG connecting the target and the origin
routers;

		   o  Along the temporary DAG created for route
discovery, provided that
		      the target router is reasonably sure that the
DAG's life time is
		      sufficiently long and the routers in the DAG do
remember one or
		      more of their parents in the DAG.

		JP> You may want to explain a bit more why you the
origin router
		would be interested in receiving an ACK without the
route itself (case
		6.3): hop-by-hop routes.
		Goyal, et al. Expires December 16, 2010 [Page 16] 
Internet-Draft draft-dt-roll-p2p-rpl-01 June 2010 6.3. DRO as Carrier of
Forward/Bidirectional Source Routes When a target router discovers a
forward source route, 

		JP> Here, I agree that this is a forward route.
		it sends a DRO
		   message to the origin router carrying the discovered
source route
		   inside a Source Route option.  Similarly, when a
target router
		   discovers a bidirectional source route, it caches the
route for its
		   own use and then sends a DRO message to the origin
router carrying
		   the discovered route inside a Source Route option.
Rather than
		   immediately sending a discovered route back to the
origin router, a
		   target router MAY optionally send one DRO message
after discovering
		   multiple forward/bidirectional source routes with the
sent DRO
		   carrying the discovered source routes (in multiple
Source Route
		   options).

		JP> Note that the algorithm for "best route" selection
is outside the scope
		of this document, so does the use of a timer on the
target side to determine
		how much time to wait before answering, ... 
		A DRO message carrying one or more Source Route options
MUST have its D field set to 0x00 (for forward routes) or 0x02 (for
bidirectional routes). The S flag MUST be set to 1 and the N field MUST
indicate the number of Source Route options in the message. A Source
Route option MAY immediately be followed by a Metric Container option
that carries the aggregated 

		JP> or recorded
		values of the routing metrics for this source
		   route.  The target router may send this DRO message
to the origin
		   router in any fashion it desires including the
options listed in
		   Section 6.2.

		6.4.  Establishing Hop-by-hop Routes Via DRO

		   In order to establish a hop-by-hop route, the target
router sends a
		   DRO message along the reverse of the discovered route
(via a Type 4
		   Routing Header specified in
[I-D.hui-6man-rpl-routing-header]).  The
		   D field of the message is set to convey the route's
direction
		   (forward/backward/bidirectional) and the S flag MUST
be clear
		   (indicating hop-by-hop).  The N field MUST be set to
1.  Such a DRO
		   MUST NOT contain any option.

		   A router receiving such a DRO message SHOULD
establish hop-by-hop
		   state in the right direction corresponding to the
route carried in
		   the Type 4 Routing Header of the DRO message.  If a
router does not
		   establish such state, it MUST drop the DRO message.
Otherwise, it
		   MUST forward the DRO message along the source route
contained in Type
		   4 Routing Header of the received message.

		JP> Any error log if it does not ? (optional) ICMP sent
to the target node?
		   A router SHOULD associate the hop-by-hop routing
state, thus
		   established, with the objective function and metrics
used for route
		   discovery.

		Goyal, et al.           Expires December 16, 2010
[Page 17]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		JP> New section to indicate how to clean-up the
temporary DAG ?
		7.  Security Considerations

		   TBA

		8.  IANA Considerations

		   TBA

		9.  Authors and Contributors

		   In addition to the editor, the authors of this
document include the
		   following individuals (listed in alphabetical order).

		   Anders Brandt, Sigma Designs, Emdrupvej 26A, 1.,
Copenhagen, Dk-2100,
		   Denmark.  Phone: +45 29609501; Email: abr@sdesigns.dk

		   Robert Cragie, Gridmerge Ltd, 89 Greenfield Crescent,
Wakefieldm WF4
		   4WA, UK.  Phone: +44 1924910888; Email:
robert.cragie@gridmerge.com

		   Jerald Martocci, Johnson Controls, Milwaukee, WI
53202, USA.  Phone:
		   +1 414 524 4010; Email:jerald.p.martocci@jci.com

		   Charles Perkins, Tellabs Inc., USA.
Email:charliep@computer.org

		   Authors gratefully acknowledge the contributions of
Richard Kelsey
		   and Zach Shelby in the development of this document.

		10.  References

		10.1.  Normative References

		   [RFC2119]  Bradner, S., "Key words for use in RFCs to
Indicate
		              Requirement Levels", BCP 14, RFC 2119,
March 1997.

		10.2.  Informative References

		   [I-D.brandt-roll-rpl-applicability-home-building]
		              Brandt, A., Baccelli, E., and R. Cragie,
"Applicability
		              Statement: The use of RPL in Building and
Home
		              Environments",

draft-brandt-roll-rpl-applicability-home-building-00 (work
		              in progress), April 2010.

		   [I-D.hui-6man-rpl-routing-header]
		              Hui, J., Vasseur, J., and D. Culler, "A
Source Routing

		Goyal, et al.           Expires December 16, 2010
[Page 18]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		              Header for RPL",
draft-hui-6man-rpl-routing-header-00
		              (work in progress), May 2010.

		   [I-D.ietf-roll-building-routing-reqs]
		              Martocci, J., Riou, N., Mil, P., and W.
Vermeylen,
		              "Building Automation Routing Requirements
in Low Power and
		              Lossy Networks",
draft-ietf-roll-building-routing-reqs-09
		              (work in progress), January 2010.

		   [I-D.ietf-roll-of0]
		              Thubert, P., "RPL Objective Function 0",
		              draft-ietf-roll-of0-02 (work in progress),
June 2010.

		   [I-D.ietf-roll-routing-metrics]
		              Vasseur, J., Kim, M., Networks, D., and H.
Chong, "Routing
		              Metrics used for Path Calculation in Low
Power and Lossy
		              Networks",
draft-ietf-roll-routing-metrics-07 (work in
		              progress), June 2010.

		   [I-D.ietf-roll-rpl]
		              Winter, T., Thubert, P., and R. Team,
"RPL: IPv6 Routing
		              Protocol for Low power and Lossy
Networks",
		              draft-ietf-roll-rpl-09 (work in progress),
June 2010.

		   [I-D.ietf-roll-terminology]
		              Vasseur, J., "Terminology in Low power And
Lossy
		              Networks", draft-ietf-roll-terminology-03
(work in
		              progress), March 2010.

		   [I-D.thubert-6man-reverse-routing-header]
		              Thubert, P., "Reverse Routing Header",

draft-thubert-6man-reverse-routing-header-00 (work in
		              progress), June 2010.

		   [RFC5826]  Brandt, A., Buron, J., and G. Porcu, "Home
Automation
		              Routing Requirements in Low-Power and
Lossy Networks",
		              RFC 5826, April 2010.

		Goyal, et al.           Expires December 16, 2010
[Page 19]

		Internet-Draft          draft-dt-roll-p2p-rpl-01
June 2010

		Authors' Addresses

		   Mukul Goyal (editor)
		   University of Wisconsin Milwaukee
		   3200 N Cramer St
		   Milwaukee, WI  53211
		   USA

		   Phone: +1 414 2295001
		   Email: mukul@uwm.edu

		   Emmanuel Baccelli (editor)
		   INRIA

		   Phone: +33-169-335-511
		   Email: Emmanuel.Baccelli@inria.fr
		   URI:   http://www.emmanuelbaccelli.org/

		   P2P Team

		JP> Remove P2P Team unless you list the people.
		Goyal, et al. Expires December 16, 2010 [Page 20]  

		_______________________________________________
		Roll mailing list
		Roll@ietf.org
		https://www.ietf.org/mailman/listinfo/roll