Ad hoc Network Autoconfiguration BOF (autoconf)
Attendees: 288
Monday, August 1 at 1630-1800
=============================
CHAIRS: Shubhranshu Singh <shubhranshu@samsung.com>
Thomas Clausen <thomas.clausen@polytechnique.fr>
Agenda
2nd Ad-hoc network Autoconfiguration BoF
Agenda
- Agenda bashing - Shubhranshu/Thomas
- Background & status - Shubhranshu
- Problem Statements by Charlie Perkins
- Proposed Charter - Thomas Clausen
- Summary of existing proposals
---------------------------------
Background & status
Shubhranshu
Four phases
Phase 1: ~ 2000
Phase 2
- several interested parties
- with indsutrial needs, deployment scenarios
Phase 3
- scenario & requirements drafts
- some implementation experiments
- large bar-BOFs, good momentum
Phase 4
- ad hoc network autoconf BoF
- IETF community interest
- comments on the work scope
Phase 4bis
- charter refinement
- IESG/IETF discussion
Current Status
- IESG approved the charter for external review
WWW-page
- http://www.lix.polytechnique.fr/Labo/Thomas.Clausen/IETF/MANET/AUTOCONF
- mailist archieve & subscription info.
---------------------------------------
Definition and Problem Statements
Charlie Perkins
draft-singh-autoconf-adp-00.txt
Background
- Ad hoc node may need to autoconfigure either or both of:
- global scope addresses
- local scope addresses
- current status
- no standard mechanism and definition related to autoconf of ad hoc nodes
- MANET list has carried discussion of autoconf ideas and requirements almost since MANET was chartered
- has never been a charter item
- several autoconf drafts
- None of them have been accepted as WG drafts
- several commercial projects have shown the need
Other WGs
- zeroconf
- took years to finish even with product experience and great interest
- warnings about having special addresses treated differently by apps
- strong requirement to prevent address from escaping the local network applications sometimes pass address references to distant computers
- IPv6 stateless autoconf
- needs deployment experience with big populations
- must consider the experience and outcome of "site-local"
- can not, natively, work in multi-hop scenario
- DHC
- DHCP is a netowrk administrative
Terminologies
- Local address: valid only within the MANET
- Standalone ad hoc networks: MANET not connected to any other network
- hybrid ad hoc network: MANET connected to infrastructured network
- internet gateway: a node which has connectivity to the Internet and enables a MANET to be reachable from the Internet (and vice versa)
- duplicate address detection
- A protocol mechanism for ensuring uniqueness of IP addresses akin to RFC2461
- A protocol mechanism for ensuring uniqueness of IP addresses in the face of network merger
Problem Statemnt 1/3
- typical feature of ad hoc networks
- Multi-hop packet forwarding
- host also serve to froward packets
- Infrastructure-less
- Random mobility
- Different concept of link
- these features require re-examiniation of existing mechanisms
- RFC 2462: stateless address autoconfiguration
- RFC 2461: Neighborhood Discovery protocol
- RFC 3315: DHCPv6
- etc
- No standard specification describing how ad hoc nodes should autoconfigure IP address and undergo DAD
Problem Statement 2/3
- A manet may appear in two different situations
- Switching between the above two
...
Problem Statement 3/3
- Network merger & partition
- inherent property of ad hoc network
- may occur at any point of time
- merger may result i naddress conflict
- relevant to standalone as well as hybrid network
Assumptions
- Nodes should be able to get IP addresses that conform to the charcteristics of the IP addressing architecture
- If a connected manet has hierarchical substructure, the addrss allocated to nodes in a subhierarchy must fit the address range associated to that subhierarchy
- Internet gateways advertising connectivity to the same routing prefix must coordinate their routing tables
- Internet gatewyas may offer several different routing prefixes
- When dpulicate addresses are detected , at least one of the nodes must discontinue
- The rpotocol should work regardless of underlying routing protocol
- But protocol features might offer significant optimization
- Lifetime for autoconfigured addresses
- If lifetime expires, use of the address should immediately
Jim Bound: One of the thing thinking is that autoconf aspect, do we assume base station in the radio or not. We need to consider that. Military, health, we need base stations.
Keath; Lots of kind of ad hoc networks. Are you trying to accomodate all the fabricts, We need to draw a line some where.
Charlie: The problem we want to look at
Discussion on Boundry Issue
Thomas: Should also be taken to the MANET WG
Keath: I don't hear IPv4 or IPv6, but hope we don't consider IPv4.
Charlie: We are not likely to do IPv4.
Thomas: IPv4 is possible, but seems more complicated. Ipv6 solution should be decent first step.
Keath: They are very different problems, and doing both might make it difficult.
Thomas: Interest in IPv4, and it might be useful to do v4 deployment. If we come up with IPv6, we can consider IPv4 later.
Thomas: Are you in an agreement to do only IPv6 now
Margaret: Current charter only considers IPv6
Dave Green: How do you address name schemes. You said DHCP is not opproational.
Charlie: This is not relevent for this discussions here.
Samita: What are the differences between hybrid and internet gateway
Thomas: Whether they can provide addresses or not
Samita: Is routing information kept inside the network or ???
Charlie: Gateway would provide some range of IP address within the MANET with routable range.
Charlie: we are not NATing.
Dave Thaler: Document looks good. PS section can be a lot better with actual problems listed. Problems are not specific to address resolution. More discussions on the detail of what the problems are in a multihop routing protocols, mutlicasting with IPv4 and IPv6. Whether it would be in this document or other references, I would like to see it. The charter does has one paragraph.
Shubhranshu: These has been discussed and next version of PS will address and calrify these issues.
Tim Sharper?: What scale do you have in mind? In scale of billions, it seems hard.
Charlie: We are to design protocols, nobody ever showed a need for billion node ad hoc network.
Thomas: The answer can be found in the protocols provided by the MANET WG. We are aimed for the scalability with the same.
Tim S: Ethernet has 6 bytes unique, are we going to use those? Where do you plan to go?
Thomas: Quick question would be that IPv6 stateless would not function correclty in MANET.
Tim S: How these fit with the trill?
Erik Normak: Its quite different
Erik N: I don't see issues with bootstrapping of MANET in the problem statement. It makes sense to put that in there.
Sam Harten?: It would be good to clarify security threat model, in which we are solving and which are not.
Jim ..: As long as there is a base station, they should work. In scalability I've seen solutions work in thousands, on simulations.
Jim : The security issue is very important since they involve scenario with medical ...
Yang: Is the addresses for the routing or identifing?
Shubhranshu: They are just like any IP address and can be used for routing and/or identification purposes.
Potential design quidelines
- A node may choose which Internet gateway's routing prefix to use for autoconf according to any convenient criterion, not neccessarily constrained by the autocof protocol
- Routes internal to the ad hoc network must not leak into the Internet
- Internet nodes cannot see past the Internet gateway
- Internet gatway can be treated as a default router towards the ...
...
---------------------------------------
Proposed Charter Thomas C.
Minneapolis
- Charter Proposal presented
- broad ideas, few firm decisions
- no milestones, deliverables, specific
- purposely open-ended - solicit/provoke thoughts
- Goals between 62 and 63
Proposed Charter
- To develop an address autoconfiguration solution, supporting
- stand-alone, hybrid and/or intermittently
...
63rd IETF in Paris
- Charter approved by the IESG for external review
- context, deliverables, milestones
- resulting from multiple iteractions ....
Autoconf Context
- MANETs assume
- no a priori infra
- multi hop environment
- Existing mechanisms for autoconf
- assume mutlicast like among nodes
- assume central entity
Deliverables
- terminology and ps
- statelsss mechanism fro IPv6/MANETs
- unique local, global (where applicable) addresses
- statefull mechanism for IPv6/MANETs
- recongnition that dhcp-hacks are used on MANETs already
- continued address-uniqueness promotion
Milestones
Oct 05: submit terminology and problem statement for WG review
Oct 05: submit intial I-D of candidate proposed AUTOCONF mechanisms and design frameworks
Feb 06: submit terminology and ps document to IESG for publication as infromational
Apr 06: submit initial ID of stateless autoconfiguration mechanism for WG review
Apr 06: submit initial ID of stateless autoconfiguration mechanism for WG review
Apr 06: submit initial ID of configured address uniqueness maintenance for WG review
2007: recharter or close
A couple of comments
- Reuse existing specification where appropriate
- e.g. we should answer "how to make IPv6 stateless autoconf or DHCPv6 work in a manet environment"
- we should not e.g. "invent a new version of the DHCPV6"
- Develop specifications:
- configured address-uniqueness
Pekka S: Is this proposed WG trying to create standard track or start with experimental. Starting with experimental might be good.
Thomas: Intention is to go for Standard track with the experieces gained already.
Keath Morgan: From the name of the WG, the scope looks larger than it intends. You might be more focused if you use the 3 bullets in the slide.
Keath Morgan: If the intension is to make standards ...
Thomas: We are not saying we are using DHCP, but there are people doing stateful mechanisms in MANET and should not see why we should not standardilize it.
---------------------------------------
Survey of IP address autoconfiguration mechanisms for MANETs
draft-bernardos-manet-autoconf-survey-00
IP autoconfiguration in MANET
- there are several protocols
- IDs and research papers
- Different scopes, targets, applicability scenarios
- IPv4/IPv6/ IP famility address independent
- Ad hoc routing protocol dependency
- support for partitioning and merging
- duplicate address detection in manets
- centralised / distributed
- for connected / disconnected MANETs
- integrated or not with Internet-Gateway
- Many of the proposals try to ...
Conflict detection allocation
draft-perkins-manet-autoconf-01
- scope
- local IPv4/IPv6 adress autoconfiguration
- ad hoc routing protocol independent
- basic mechanism
- a node chooses an address randomly and performs DAD within the MANET
- DAD within the MANET
- Based on sending AREQ messages and waiting for AREP messages
- Smae mechanism for IPv4/IPv6
draft-weniger-manet-addressautoconf-ipv6-00
- scope
- IPv6 address auto-configuration for MANETs
- ad hoc routing prootcol independent
- basic mechanism
- A hierarchy is established by special nodes that configure a group of nodes by issuing modifed RA within their scope
- DAD is performed within the scope of the node
- The subnet ID has a unique...
draft-wakikawa-manet-globalv6-04
- scope
- global IPv6 address autoconf for hybrid MANETs
- ad hoc routing protocol independent
- basic mechanism
- exists at least one internet-gateway that sends proactively or reactively I-G advertisements that contain prefix infromation
- extended RS/RA messages
- a node configures a global IPv6 address from...
draft-jeong-adhoc-ip-addr-autoconf-04
- scope
- local IPv4/IPv6 address autoconfiguration for disconnected MANETs
- ad hoc routing protocol independent
- Considers partitioning/merging
- Basic mechanism
- address autoconfiguratio ncomprised of three steps
- selection of random address
- verfication of the address uniqueness
- assignment of the address to the interface
- Besides strong DAD during initailisation, intermediate routers also check for address duplication
draft-ruffino-manet-autoconf-mutigw-00
- scope
- global IPv6 address autoconfiguration for hybrid MANETs
- specified for OLSR, but can be generalized
- basic mechanism
- there exist several gateway available in the MANET. Each gateway has a global IPv6 prefix that is announced using a new OLSR message type (PA)
- at bootstrap, a node configures a PADD as main address in OLSR
- with the prefix information from PAs, a node is able to build a set of global IPv6 addresses
- a generic DAD procedure should be taken...
draft-laouiti-manet-olsr-address-autoconf-01
- scope
- DAD for MANETs
- Specified as an extension to OLSR
- support partitioning/merging
- basic mechanism
- Each node in MANET includes it IP addresses
....
draft-singh-manet-mmg-00
- global IPv6 address autoconf for hybrids
- ad hoc routing protocol independent
...
draft-mase-manet-autoconf-noaolsr-00
scope
- IPv6 address autoconfiguraiton for MANETs
- modification of the OLSR spec
- basic mechanism
- 3 parts
- address selection
- onggoing DAD using...
Conflict-free allocation
draft-clausen-manet-address-autoconf-00
- scope
- global IPv6 address autoonf for hybrids
- specified for OLSR, but can be generalized
- basic mechanism
- there exists at least one configured node in the MANET
- the configured nodes periodically beacons
- a new node slects one configured
...
draft-jelger-manet-gateway-autoconf-v6-02
- scope
- global Ipv6 address autoconfiguration for hybrids
- ad hoc routing protocol independent
- basic mechanism
- each gateway present in the MANET sends periodically GW INFO messages containing IPv6 global prefix information to its one hop neighbors
- each node in the MANET selects one prefix for the configuration of its address
- DAD not performed
non I-D proposals
- there are severl proposals published as papers and journals
- some introduce concepts and mechanisms that are afterwards reused by some of the IDs
- Interesting conflict-free allocation approaches
- refernces to these papers can
Heather?: Instead of organization different solution, how these solution ...
Thomas: permechanism servey
Erik N: Some of the solution assume EUI64.
Erik N: Are there one problem or two? Disconencted ad hoc maybe a different problem as supposed to hybrid case which is a natural way to provide address.
Thomas: I think you are right. With infra it would make things easier. Point is that network might be connected but then get disconnected
Charlie P: Experience up until now, a single mechnism can solve both of the problems. One of the problem is that some of the wireless devices does not know if its disconnected or not. We should use the experience and proceed.
---------------------------------------
Intersting topic for the IETF; should support? (Hmmm....)
many
Not intersting topic for the IETF; should support?
None.
How many think we should charter wg with the current charter
many
How many think we should not charter wg with the current charter
~2
|