Re: [Autoconf] Closing summary on consensus-call for RFC5889modifications

[Teco Boot <[e-mail]>]

Hi Charlie,

More thoughts on the grey zone, separating host and routers.
You stick to the basic definition: routers forward packets, hosts do not. 
OK?

We had discussion on routing protocol speakers. I said: sending routing 
protocol packets makes a node a router.

But for routing protocol speakers, that:
 - do not forward packets
 - do not send RAs
 - do not advertise any willingness to forward packets
Would that be routers?

If you say: no, these are hosts, I can follow.
But for accepting such, it should be made clear to other nodes, that 
the routing protocol speaker is a host, not a router.
Some mechanisms:
 - passive mode / don't send routing protocol packets at all
 - OLSR willingness = WILL_NEVER (0)
 - not advertise any links
 - advertise links, but with "do not use" attribute / metric
 - only put reachability for itself in AODV / DYMO messages
 - never send BGP reachability with own address as next_hop

Using this classification, a BGP route server is a nice example for a 
host being an active routing protocol speaker.
draft-jasinska-ix-bgp-route-server-00:
|  Although a route server uses BGP to exchange
|  reachability information with each of its clients, it does not
|  forward traffic itself and is therefore not a router.

If we agree on such classification, there is some advertorial work to do.
Because many would say (ad hoc) routing protocol speakers are routers. 
E.g. we need errata on RFC's from MANET WG and update the active drafts.
Also, draft-ietf-autoconf-adhoc-addr-model-03 needs a revision. Change 
/router/node/, but not global change (not in section 6.1, last bullet).
RFC5998 is waiting for approval by Mark. 
Mark?

And using Router-ID's on hosts?
Hmmm.

Regards, Teco



Op 26 aug 2010, om 20:17 heeft Teco Boot het volgende geschreven:

> Charlie,
> 
> Learning default gateways in a passive mode is well accepted behavior
> of hosts. It is mentioned in RFC 1122:
> |  This technique depends upon the host passively
> |  receiving ("wiretapping") the Interior Gateway Protocol
> |  (IGP) datagrams that the gateways are broadcasting to
> | each other. 
> 
> But signaling from hosts to routers is another story. We have NDP or 
> ARP for this. One could think of redistributing NDP / ARP learned routing
> info into the routing protocol. I am not sure we should go there.
> 
> Teco