Re: [DNA] Re: RS/RA Exchange

[Greg Daley <greg.daley@eng.monash.edu.au>]

Hi Erik,

Erik Nordmark wrote:
>>How about an RA option that lists all "learned" peers?
> 
> 
> I think you want to put any additional information in packets that
> are only sent between the routers. The RAs are sent to all-nodes hence
> the hosts will see all these added bytes which they don't derive any use from.

There's some ideas here which may have parallels with VRRP.
(private - routers only signalling, effects on router
discovery/forwarding).

Personally, I think that the lack of security in VRRP is the
killer for using this technology in DNA.  It may be trivial
on a wireless link to become the [fastest responder, VRRP master].

I'd guess that even if we're only sending to routers, it's
best to used router discovery with SEND.  The delegation
mechanisms in SEND have at least some chance of implying
right to route.

I've been playing around with this sort of signalling
for another purpose where router ordering is important.
After discussions with Brett here, I guess the
best mechanism is to explicitly indicate interest or
membership in an ordered set of devices (possibly
communicating that they have a particular idea of the set),
and devices agreeing on a nominal order between themselves.

Depending on the application (RA response ordering, link
identification, etc), there may be different requirements.
I'm trying to look at a subset of these to see if there's
common ground.

There are still problems in any wireless environment
with VRRP or router-to-router discovery when L2 switches treat
multicast like broadcast since the configuration packets
will all travel onto the constrained wireless link, but at
least hosts don't have to wake up to process the packets on
the wireless link.

Greg