Re: [DNA] DNA Goals Issue List

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

----- Original Message -----
From: Erik Nordmark <Erik.Nordmark@sun.com>
Date: Wednesday, August 4, 2004 8:43 am
Subject: Re: [DNA] DNA Goals Issue List

> > I am not sure you need to detect the identity of currently 
> attached link for
> > quick configuration either. For example, considering the CPL as 
> the link
> > identity, you don't  need to detect the identity of currently 
> attached link
> > (the COMPLETE prefix list) to ascertain the validity of the 
> existing IP
> > configuration or to quickly re-configure  yourself. You only 
> need one prefix
> > - right? So, making "detecting the identity of  the currently 
> attached link"
> > part of the goal is kinda misleading.
> 
> If the host is on the same link as before, it needs at least on 
> prefix which
> was part of the known set of prefixes for the link.
> (Note that due to packet loss of RAs a host might not know all the
> prefixes on the attached link.)
> 
> If the host has moved to a different different link, it needs to see
> enough RAs and prefixes to have a pretty good probability that the
> prefix sets is disjoint.
> If it knows the complete prefix list from the old link with high 
> probability,then one RA might be sufficient.
> 
> There is a very unlikely case where this isn't sufficient, which 
> is when the
> host moves between APs (i.e. receives a 'link up' indication from 
> the device
> driver) and *at the same time* a new router is added on the link 
> and a new
> prefix is added, and the new router only announced the new prefix, 
> and the RA
> from this new router arrives at the host first after the host has 
> sent the RS.
> In that case, from the single RA with the single prefix, it will look
> like a different link because the prefix sets are disjoint.
> But this is very unlikely.

Indeed.

In some situations, it may not be worth worrying about,
in which case we get one spurious configuration.

If we state this in a solutions or CPL document then
that's probably OK.

Greg
 
>  Erik
> 
>