Re: [DNA] Best current practice for DNA

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

Hi Syam.

----- Original Message -----
From: Syam Madanapalli <syam@samsung.com>
Date: Monday, March 1, 2004 6:56 pm
Subject: Re: [DNA] Best current practice for DNA

> Hi Greg,
> 
> ----- Original Message -----
> From: "Gregory Daley" <Greg.Daley@eng.monash.edu.au>
> To: "Syam Madanapalli" <syam@samsung.com>
> Cc: <dna@eng.monash.edu.au>; "Youn-Hee Han" <yh21.han@samsung.com>;
> "JinHyeock Choi" <athene@sait.samsung.co.kr>
> Sent: Monday, March 01, 2004 1:47 PM
> Subject: Re: [DNA] Best current practice for DNA
> 
> 
> > Hi Syam,
> >
> > ----- Original Message -----
> > From: Syam Madanapalli <syam@samsung.com>
> > Date: Monday, March 1, 2004 4:29 am
> > Subject: Re: [DNA] Best current practice for DNA
> >
> > > Hi Greg,
> > >
> > > How about "Movement Detections" that are in use currently and 
> their> > evaluation,
> > > that we have mentioned in the following document for BCP?
> > >
> > > http://www.thinkonweb.com/sait/MD.pdf
> >
> >
> > This looks like a good starting point for the
> > analysis (at least in checking Router reachability).
> 
> As of now I think most of the people use Router
> Reachability check for detecting that it has moved.

I'm not sure which one though: 1NS,NUD,RS/RA?

even using RS/RA has difficulties.

Non-send nodes cannot identify multicast responses
to RS as directed to themselves (or even responses). 

It may be that unicast RA responses are sufficient 
in this case, but I've still to complete an analysis 
of the overhead associated with this.
(Not that it's hard to do, it's just not been looked
at deeply I guess).


> >
> > I guess that there may be some differing circumstances
> > depending on the address state for DAD, though
> > For example, (non DAD) NS's while still in tentative state may
> > be difficult.
> 
>  This is true. Currently we use the existing LLA (Link Local 
> Address).So I am wondering what if we do DAD for Router Address to 
> checkits reachability. If the router reachable (we did not move), 
> the Router
> will defend by sending DAD NA.

While this will work in some cases, the ambiguity of
the link-local router address is 

Also, in SEND, it may be that the DAD NS is ignored. 
Since there is only one CGA to bind the address to (the Router's
target CGA).  As we don't have the private key associated
with the CGA, we cannot generate a signed SEND message
using the public/private keypair used with the CGA.

It may be that the DAD attempt is ignored if the router
isn't tentative (if the router interprets it as an attack).
 
> >
> > If we believe that some of these ideas aren't
> > generally applicable, or have inappropriate features
> > we can reduce the solutions to the 'best current'.
> 
> We talked to the people in ETSI IPv6 IOP in Brussels
> last year; based on this we have developed these MD
> Schemes and evaluated. Probably we can collect more
> information on the mailing list and try to modify this document.

That sounds like a good idea. 

Certainly looking at existing MIPv6, ND/SAA or DHCPv6
implementations and what they already do would be a good
starting point.

Greg