Re: [Autoconf] Call for comments to a new AUTOCONF charter proposal.

[Alexandru Petrescu <[e-mail]>]

Le 29/06/2010 08:03, Henning Rogge a écrit :
> On Tue June 29 2010 06:20:28 Ryuji Wakikawa wrote:
>> Description of Working Group:
>>
>> RFC 5889 presents one possible IPv6 addressing model for ad hoc nodes. In
>> this model the ad hoc routers need to configure their network interface(s)
>> with addresses valid in the ad hoc network, and may configure additional
>> prefixes for use by attached nodes.
>>
>> After completing the work on RFC 5889, the main purpose of the AUTOCONF WG
>> is to standardize how existing IPv6 address configuration tools can be
>> used for address configuration:
>>
>> 1. A DHCPv6-based mechanism for configuring required interface addresses
>> for the routers in the ad hoc network. This mechanism is expected to
>> produce addresses with properties outlined in RFC 5889. This mechanism
>> uses the existing DHCPv6 protocol unchanged, and assumes a central node
>> that can allocate addresses on a first-come-first-served basis. Other
>> nodes in the ad hoc network will relay messages to the central node in
>> order to help a new node get an address for itself. This mechanism is only
>> suitable for deployments were a central node can be set up. It should be
>> noted that many existing deployments employ Internet gateways that can act
>> as such a central node as well. Future extensions such as central node
>> election may make this mechanism suitable for also for stand-alone ad hoc
>> networks.
 >
> I don't think elections are a good way to add redundancy for an address
> configuration system. The dependency of a single node in the whole network to
> configure the addresses will be a bottleneck for larger mesh networks.

But elections is what gets rid of central node bottlenecks... if that 
single node fails then elect another.

>> 2. A DHCPv6-based mechanism for delegating a prefix(es) to each router for
>> use by applications running on the routers themselves, or for
>> configuration of attached hosts/networks. This mechanism works in a
>> similar manner to the one above, but allocates prefixes instead of
>> addresses.
>
> I don't see a difference between 1 and 2, for IPv6 case 1 is just case 2 with
> a prefix length of 128 (or 64 if we use the IPv6 autoconf postfix).

To my quick reading 2) does prefixes too whereas 1) does addresses only.

Alex

>
> --------------
> Does this proposal mean the autoconf group will not work on a distributed
> address configuration scheme for mesh networks ?
>
> Henning Rogge
>
>
>
> _______________________________________________
> Autoconf mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/autoconf