2.3.2 Ad hoc Network Configuration (autoconf)

NOTE: This charter is a snapshot of the 62nd IETF Meeting in Minneapolis, MN USA. It may now be out-of-date.

Last Modified: 2005-02-09

Chair(s):

Thomas Clausen <thomas.clausen@polytechnique.fr>
Shubhranshu Singh <shubhranshu@samsung.com>

Internet Area Director(s):

Thomas Narten <narten@us.ibm.com>
Margaret Wasserman <margaret@thingmagic.com>

Internet Area Advisor:

Margaret Wasserman <margaret@thingmagic.com>

Mailing Lists:

General Discussion:
To Subscribe:
Archive:

Description of Working Group:

A group of mobile wireless nodes capable of spontaneously forming
network and support multi-hop communications constitutes a mobile ad
hoc network (MANET). One of the main reason for popularity of manet is
that it does not rely on any existing infrastructure.  Several manet
routing protocol specifications have been developed by IETF's MANET
WG. However, currently there is no standard specification that can be
used by manet nodes to auto-configuring its IP address and other
network parameters. For IP-based infrastructure network there are
several specifications for this purpose such as RFCs 2461, 2462,3315
etc. However, considering the unique properties of ad hoc networks
such as multi-hop packet forwarding, absence of infrastrastructure
e.g. DHCP server, these specifications cannot be directly used by
nodes constituting ad hoc network. 

The main goal of the MANET AUTOCONF is to develop solutions for IPv4
and IPv6 address auto-configure that are either manet-local scope i.e.
valid only within a particular ad hoc network or global scope i.e.
addresses that are routable on the global Internet.   
   
Stand-alone ad hoc network:   
Such ad hoc networks are not connected to any external network e.g.
Conference network, battlefield network, surveillance network, etc.
For such manet, solutions for IPv4 and IPv6 address auto-
configuration, duplicate address detection and resolution will be
developed considering situation where nodes may be added or removed
randomly. Any address autoconfiguration mechanism for such networks
should take care of the lack of any pre-established and reliable
address or prefix allocation agency.   
 
Ad hoc network at the edge of infra-structure network:
Stand-alone network connected to the Internet via one or more Internet
gateways i.e nodes sitting in between the two networks comes under
this scenario. Internet gateway may be either fixed or mobile, single
or multiple, equipped with wired and/or wireless interfaces and should
be able to provide global addressing as well as bidirectional
connectivity to those manet nodes connected to it either directly (1-
hop) or through one or more intermediate nodes. Manet nodes may use
Internet gateway for prefix allocation. Presence of Internet gateway
may make address configuration mechanism comparatively simple but
introduces other issues such as how manet nodes receive or Internet
gateway provides topologically correct routing prefix to the manet
nodes, how to select gateway, etc. Hence, for such network sufficient
detail about Internet gateway will be provided, along with address
autoconfiguration solution.
   
The third scenario where ad hoc network may be stand-alone for most of
the time but temporarily connected to the infrastructured network e.g.
car network connected while parked and disconnected otherwise, will
also be considered.     
   
Two or more ad hoc network may join together or single ad hoc network
may get partitioned into two or more separate networks, at any moment
of time. While network partitioning may not cause any severe problem
to the node functioning, network merging may create problem by
introducing more than one node in a MANET with the same address.
Hence, appropriate mechanism should be provided in order to ensure
uniqueness of the configured interface addresses after network merger.
Any other problem arising due to network merger and partitioning
should also be taken appropriate care.

Goals and Milestones:

No Current Internet-Drafts

No Request For Comments

Current Meeting Report

Monday, March 8th 2005 at 0900-1130

===========================================================================

Chair(s):

Thomas Clausen <T.Clausen@computer.com>

Shubhranshu Singh <Shubhranshu@samsung.com>

Scribe:

Emmanuel Baccelli <Emmanuel.Baccelli@inria.fr>

===========================================================================


Agenda:

- Agenda Bashing (Shubhranshu/Clausen) - 05 Minutes

- Background & Status (Shubhranshu/Clausen) - 15 Minutes

- Summary of existing proposals (Shubhranshu/Clausen) - 20 Minutes

- Connectivity Scenarii for Manet (Ruffino) - 15 Minutes

- Problem Statement (Perkins) - 20 Minutes

- Proposed Charter (Shubhranshu/Clausen) - 5 Minutes

- Open Discussion - 30 Minutes

- Should a WG exist?


============================================================================


- Agenda Bashing (Clausen)

No comments


- Background & Status (Clausen)

(see slides)

No comments


- Summary of existing proposals (Shubhranshu)

(see slides)


Question: any problem statement existing? Doubt over the fact that only one solution to everything can be produced.

Clausen: yes, many drafts contain requirements specifications, or scenarii descriptions. There is also an existing tentative problem statement draft. We intend to update this draft with the feedback we get in general after this bof.


Other comment: The result of this work has to be one and only one solution
(for the industry to be able to use this work)


- Connectivity Scenarii for Manet (Ruffino)

(see slides)


Question: What is manet specific and what is more generic problems?

Ruffino: The difference is the multi-hop aspect. Standardized autoconf mechanisms don't work with that.

Clausen: Another question is: what is a link in the wireless environment? We have to reach a consensus on this. This is a requirement to go further.

Ruffino: The partionning/merging aspect is also specific to manet.


Question: Is this a L2 or L3 problem? Do you decrement TTL at each hop?

Clausen: Yes we do, so this is a L3 problem, for now at least.

Question: Are transit networks envisionned?

Clausen: we are not considering this as in the scope.

Margaret: It is not clear to me how you run IP over the multihop wireless scenarii described. Is this IP routing without the subnet philosophy?

Clausen: See MANET work: basically nodes are both end-points and routers.

Margaret: It is still not clear to me how to break free from the subnet philosophy.

Clausen: there is no aggregation done, only flat routing.

Charlie: Link definition must be addressed. Layer 2, or layer 2.5?

Question: We have to understand why a transit ad-hoc network and an edge ad hoc network is different. Has the Manet produced a document about the architecture of ad hoc network? With the definition of an IP link in this environment? We need an agreement on this.

Charlie: There are several different opinions on this problem, as on various other things.

Question: What is a gateway? Does it have to be a router, does it have to be connected to the internet?


Clausen: From a MANET perspective: yes it is a router. It may not be conected to the internet.

Question: Why not do prefix delegation, since every node is a router and/or a gateway?

Clausen: yes this might be possible, but since the nodes are mobile, we have to solve the problem of how to do this properly.

Question: Is the scope of this BOF more than doing auto-conf? Is configuring security parameters in scope?

Clausen: It's a good point, we should take it to the mailing list.


- Problem Statement (Perkins)

(see slides)


Question: In case of disconnected networks, it may be causing problems to keep previously assigned prefixes?

Perkins: Individual nodes should be able to maintain their address. A node coming in contact with a new gateway will possibly get a new prefix, and from then on the prefix may change

Question: Is this work (autoconf) a routing problem or not?

Perkins: it is hard to distinguish, but i think it is not a routing problem.

Question: Take care of ongoing sessions when doing DAD?

Clausen: it is a good point but out of the scope here.

Question: Is there any particular reason you think hierarchy would be good? How would DAD relate to that.

Perkins: yes, hierarchy is good for scalability and aggregation reasons. DAD would still be distributed.

Question: Do we consider to use MIPv6 and MIPv4 to route to a node with permanent address that moves?

Clausen: yes. We need to use such mechanisms, and probably NEMO too.

Question: If gateways are ad hoc nodes we need to be careful about this case, relatively to the routing protocol in use. Also, Mobile IP might not be a good solution for all types of situation and nodes, might be too heavy.

Clausen: the way IETF deals with moving nodes/addresses is Mobile IP. You could try to extend it, but you have to go through it if you want to have an IETF mechanism.

Question: Electing a leader in the manet and use that for autoconf and connectivity outside the manet is good.

Perkins: agreed. Lets discuss that.

Question: Not all nodes want to route packets. This should be explicitly allowed and taken into account.

Perkins: I agree with that, sensor networks are good example for that.

Question (CRC): About hierarchy, are we talking about routing hierarchy or address hierarchy?

Perkins: routing is not the scope here, but we want to enable address hierarchy (this is the point).

Question: About the merging/splitting, can we distinguish the routing and the addressing problem?

Perkins: we think it is possible to distinguish and we want to address the addressing problem.

QUestion: Manet routing: multihop neighbor discovery. Can we be independent from the routing protocol in place? What are we assuming?

Clausen: We don't want to design a new routing protocol. We have to seperate from routing scope.

Question: What are we configuring? Only routers?

Clausen: yes , we are configuring nodes that run the protocol. But these nodes might also ask for a whole range of address to itself configure nodes behind him that don't run the protocol.

Question: We need to consider if DNS related problems are in scope here

- Conclusion about wether or not to create a WG (Clausen)


Comment: If we don't want to change the IP model of subnet, the goals are doable in a 5 pages document, wether it's layer2 or layer3.

Question: Are we going to refer to MANET documents?

Clausen: it depends on the timing. We can't reference non standard tracks.

Question (AD): There is WG matter. But the problem is to broad. We need to focus more on a specific problem

Clausen: Agreed, we need to focus more for the charter.

Chris: Auto-configuration is a real requirement, and we need to address it.

Ian: a group would be appropriate. But we need to have a clear goal.

Margaret: Do we have a scoped problem that is not research but IETF work? What would be the deliverables exactly?

Clausen: (showing the slide of the tentative proposed charter)

Charlie: I don't think this work would break IP model, at least not more than Mobile IP. We don't have to solve the problem of address permanence and routing in this work.


Comment: It is unclear that we are NOT going to break IP. What is the model for a link? If we model it as L2 and IP on top it's ok. But if not, we might break IP.

Alex Zinin: I don't think MANET breaks IP. It is a challenge but it does not. We have documentation on MANET.

Margaret: The use of the term MANET is confusing, what do you mean by Mobile Ad-Hoc Network.

Slides

Agenda
Summary of proposals
Connectivity Scenarios
Problem Statement
Charter?