[manet] Re: MANET WG Charter and MANET Internetworking

Juliusz, thank you for bringing this information forward. I understand the motivation
but when we configure an Overlay Multilink Network Interface on each MANET router
we can use standard and unmodified IPv6 autoconfiguration protocols (IPv6 ND per
RFC4861/2 and DHCPv6 per RFC8415) since all MANET routers are single-hop
neighbors on a shared NBMA link. It would work the same way for any MANET
routing protocol (including your BABEL system) and take advantage of widely-
deployed standard protocols - I use unmodified kea and radvd in the reference
implementation I am building.

So, thinking of your BABEL where all nodes configure a statistically-unique 8-octet
router ID, if the border node were configured to send RA messages over the OMNI
link with PIO flags set to A=1; L=0 you could do true SLAAC and just not consider
the prefix on-link. DAD would be optimistic per RFC4429, and it should all just work.
You could then even register the SLAAC-configured address with the DHCPv6
server to make a stateful registration of the address.

Thank you - Fred

> -----Original Message-----
> From: Juliusz Chroboczek <[email protected]>
> Sent: Thursday, July 17, 2025 12:53 PM
> To: Templin (US), Fred L <[email protected]>
> Cc: manet <[email protected]>; [email protected]
> Subject: [manet] Re: MANET WG Charter and MANET Internetworking
> 
> >>> The two autoconfiguration alternatives are 1) stateless based on SLAAC,
> 
> >> SLAAC doesn't work with routers.  We've all tried, but it cannot possibly
> >> work.
> 
> > I used the term SLAAC in an imprecise way.
> 
> Ah, okay then.
> 
> > Can we conceptualize an 8+8 address autoconfiguration scheme where
> > the only "state" is the MANET router's unique interface identifier
> 
> Not only can we conceptualise it, I've actually implemented it back in
> 2007.  Unfortunately, I then changed everything, in order to make it into
> a stateful protocol:
> 
>   https://datatracker.ietf.org/doc/draft-chroboczek-ahcp/
>   https://www.irif.fr/~jch/software/ahcp/
>   https://github.com/jech/ahcpd
> 
> I could probably recover the original stateless implementation.
> 
> 
> > This information could be propagated to all MANET routers by the MANET
> > routing protocol.
> 
> Please, no.
> 
>   https://datatracker.ietf.org/doc/draft-chroboczek-homenet-configuration-separate/
> 
> >>> Here again, an overlay multilink network interface is needed to
> >>> coordinate the multiple interfaces.
> 
> >> I must have missed a step.  Why is that required?
> 
> > To give all MANET routers the appearance of being on-link with the current
> > set of Internet access points and each other even if they may be multiple
> > hops away in the MANET underlay. The alternative in the DHCPv6 case is to
> > have all MANET routers also act as DHCPv6 relay agents, but that might not
> > work so well in dynamic MANETs with rapidly changing topologies.
> 
> I don't agree that it's required.  It might, however, be a good idea,
> since it provides better layering than the alternative.
> 
> -- Juliusz
> 
> _______________________________________________
> manet mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
_______________________________________________
manet mailing list -- [email protected]
To unsubscribe send an email to [email protected]