Autoconf WG Thursday 22nd March 2007 13:00

------------------------------------------



WG milestones: slipped on February 2007 submit MANET

architecture document. May impact on following milestones.



MANET architecture document: not tutorial (previous meeting).

-01 draft, feedback good. Suggest WGLC.



Charles Perkins: some editorial changes needed, suggested

WGLC on -02, which should follow shortly.



Thomas Narton: suggested some reviewers outside MANET

community, he would be one, suggested INT area for candidates.



Problem Statement (presented by Emannuel Baccelli) updated

from -01 to -02, including removal of diagrams and alignment

with architecture document.



Problem is to provide MANET-wide unique prefixes (up to /128)

to each MANET router, if gateway is present to provide unique

global prefixes and detect and resolve non-unique prefixes

(e.g. after merging).



Alex Petrescu should this include protocol reuse where appropriate.



Jari Arkko: reuse does not belong in Problem Statement.



Simone Ruffino: suggested modifications needed, including

multiple gateways and security.



Thomas Clausen: this is individual submission, need a WG

document. No alternatives, but should the WG adopt? The

previous points agreed by authors.



Hassnaa Moustafa Suggested nothing special for security in multihop

networks. Emmanuel Baccelli suggested multihop problem not solved,

what standards were there to solve it?



Charles Perkins: Draft uses "continuous" with regard to address

uniqueness, and this should not be a requirement. Emmanuel

Baccelli suggested only problem is continuous. Text found and

says "continuous checking". Agreed document need



Hakima Chaouichi are first and second requirments same? No,

because MANET may or may not be connected to Internet. Thomas

Clausen: is document clear on these scenarios? Are there cases

where do not need full version of these? Simone Ruffino: should

we derive requirements from scenarios?



Ryuji Wakikawa: Support as WG document. Internet gateway may

have specific requirements. Is it a MANET node or not? Emmaunuel

Baccelli, part of solution space. Ian Chakeres, move away from

Internet gateway towards server able to provide what is needed.



Simone Ruffino: Internet gateway is MANET border router. Also runs

MANET protocol.



Take discussion to mailing list. Simone Ruffino: should solutions

solve all three problems. Thomas Clausen: one or more solutions

should fall out later. Not mandated if not needed. Whether modes

of a protocol, or multiple protocols part of design. Emmanuel

Baccelli, document is coherent, but can change terminology.

Need more people to have read document. Next version may become

WG document.



Fred Templin individual submission. MANETs are multilink sites,

link-scoped functions do not work.



Charles Perkins: Is it correct that MANET interface does not

have global addresses? Register objection.



Two solutions: multilink, or virtual Ethernet using IP-in-IP

encapsulation. Latter idea is not new. MANEt appears as

unified link, TTL not decremented. Standard ND works as

normal for example. No pre-service DAD, passive in-service

DAD. Three IDs.



Christopher Dearlove: Do you need multicast? SMF desirable.

Does this add overhead? Can route data traffic simply, so

apart from header overheads, no.



Alex Petrescu: Do we need virtual interfaces? Answer if we want

to use standard ND, yes. Disagreed.



Charles Perkins. Architecture document permits simply using

same addresses without virtual interfaces. Does the virtual

interface require all devices to participate? Yes. Alternative

is multi-hop ND. Would need standardisation. Disagreed on

which was preferable.



Roberto Baldessari: Suggested MANET interfaces would have

multiple addresses.



Simone Ruffino: Question about names.



Thomas Clausen. This pre-dates architecture document. Does

it match it? Yes.



Ryuji Wakikawa presented MANEMO. Stated with examples from

RFC 3753 Mobility Related Terminology. Then summarised NEMO.

Ian Chakeres, NEMO MR does not run a routing protocol as

part of NEMO protocol. Path may be non-optimised, improvement

needed.



Dow Street: Why do we have to solve this problem in NEMO? It's

a routing problem. Ryuji Wakikawa yes. Should problems be taken

into wider routing community? Ian Chakeres elicited confirmation

that MRs may be multiple hops apart. Fred Templin, can't a

MANET routing protocol solve problem?



RFC 3963 address configuration in nested NEMO does not need

MANET Autoconf. Fred Templin suggested address reuse problem,

Ryuji Wakikawa indicated addresses discarded. Thomas Clausen

notes an assumption that an interface knows when it connects

or disconnects. This limits use. Thomas Clausen (as participant)

disagrees with that this is best approach. Alex Petrescu noted

that have multiple implementation working software. Fred Templin

noted that still need applicability statement to indicate which

cases work. Teco Boot: NEMO based on Mobile IP. If really a

problem, would be for Mobile IP. Fred Templin, in MANET ingress

and egress interfaces would be same interface. Thomas Clausen,

this discussion digressing from Autoconf.



Thomas Clausen: Nodes start with globally reachable address,

but now want a topologically correct address. Is this as the

Autoconf Problem statement point 2? Qualified yes. Alex Petrescu,

problem is very similar, but this is not the whole of the story.

Ryuji Wakikawa: Current status is discussing problem.



Dow Street: Much work is treating various mobility problems as

special cases. Need to engage in wider process.



Thomas Clausen: Request MANEMO people to study architecture and

Autoconf Problem Statement documents.



Carlos Bernardino: Will be updating draft covering solutions.

Input requested.



Shubhranshu Singh summed up.